URSA Mini Pro 4.6K G2 vs. G1 Dynamic Range and Latitude Test

January 9th, 2020
URSA Mini Pro 4.6K G2 vs. G1 Dynamic Range and Latitude Test

When the new URSA Mini Pro 4.6K G2 arrived at the cinema5D office this summer, we were quite curious to see if the lab results would show a difference as Blackmagic Design claims in their official press release that, “the second generation URSA Mini Pro features fully redesigned electronics and a new Super 35mm 4.6K HDR image sensor”. Also, we wanted to see how it fares in our latitude test versus the first generation URSA Mini Pro 4.6K G1. Curious to see the results? Read on…

Last year when we tested the predecessor to the current G2, the UMP 4.6k G1 in ProRes and BRAW, we (back then) got the second-highest dynamic range result ever: 12.5 stops in ProRes, 12.6 stops for BRAW (for a signal to noise ratio of 2). See our article from December 2018 here.

So, is the new G2 going to set a new standard in our lab? Read on for our results – also, please have a look at how we test dynamic range here.

The test was performed in June 2019 – so I have to apologize to you, our readers that I didn’t publish it earlier. But, better late than never. Also, back this summer we hadn’t established our rolling shutter test yet, so this value is, therefore, missing in our results.

Before the dynamic range test, I let the UMP G2 warm-up for 15 minutes and then performed the “black shading” sensor calibration with the lens cap on. The firmware version was 6.4.

4.6K G2 ProRes XQ ISO 800 Dynamic Range Results

As a first step, I analyzed the 4.6k ProRes XQ dynamic range. At ISO800 the UMP 4.6K G2 comes in with a strong 12.6 stops of dynamic range at a signal to noise ratio of 2 according to our IMATEST analysis (13.5 stops at a signal to noise ratio of 1). This is almost exactly the same result that I got with the previous generation UMP 4.6K in December, within 0.1 stops.

IMATEST dynamic range result of the URSA Mini Pro 4.6K G2 using ProRes XQ at ISO800. 12.6 stops are calculated for a signal to noise ratio of 2.

The waveform plot below of the XYLA21 stepchart visually reveals about 12 stops above the noise floor. As can be seen, the noise floor looks quite clean, hinting at some form of noise reduction.

Waveform plot of the XYLA21 stepchart URSA Mini Pro 4.6K G2 using ProRes XQ at ISO800. 12 stops can be identified above the noise floor.

4.6K G2 BRAW Dynamic Range Results

Now, there is a little issue with BRAW and our standardized testing workflow, as we usually try to avoid any 3rd party software when extracting keyframes for our IMATEST analysis (other than ffmpeg) to avoid any kind of manipulation/influence of this software to the image. With BRAW this philosophy doesn’t work, as we have to use DaVinci Resolve to develop the files.

ISO 800 BRAW constant bitrate 3:1 4.6K scaled to UHD

Now, using ISO800 yielded 12.1 stops at a signal to noise ratio of 2 (13.2 stops for SNR = 1). Quite interestingly, this is about half a stop lower than the result we got on the UMP 4.6K tests in December last year.

It seems that BLACKMAGIC decided for BRAW on the G2 to leave the image relatively untouched (whereas in ProRes a slight internal noise reduction seems to be present in the files – this is also the case for the BRAW UMP 4.6K files from last year).

The same situation was found with the BMPCC4k when we initially tested it – also here, the ProRes files showed a higher DR than the RAW (CDNG in our tests).

As our intention is to test the untouched files straight off the camera, we are not processing the BRAW files further, hence we will use the ProRes result instead.

In general, as more and more raw capable cameras are hitting the market, we have to think about how we should test the raw modes. Of course, you can push around the BRAW files in post heavily – using noise reduction, for example, cleans up the lower stops nicely and highlight recovery would be another possibility to squeeze out even more dynamic range.

For now, we rate the URSA Mini Pro G2 4.6K with a dynamic range of 12.6 stops at a signal to noise ratio of 2  (as its predecessor, the UMP 4.6K).  

Latitude test of the URSA Mini Pro 4.6K G2 versus G1

Latitude, not to be confused with dynamic range, is the capability of a camera to retain color and detail when over- or underexposing the image. You can also view it as the margin a camera allows for exposure error.

Hence, when testing dynamic range we basically set a threshold for acceptable noise (the signal to noise ratio of 2) whereas for latitude tests we judge how well a camera retains the colors and overall image quality. Which is subjective of course, as we do not have a mathematical procedure behind it (like IMATEST for the DR measurements).

Latitude test setup with the URSA Mini Pro G1, G2, and the Panasonic S1 which we tested in the same run as well (you can directly compare the results to the URSA Mini’s by clicking here)

Studio Setup

For this test, we set up our cameras at the cinema5D studio as depicted above. We managed to put the UMP 4.6K G1 and G2 cameras quite close side by side.

Camera Setup and color grade:
Using waveform monitors, the skin (Johnnie) was exposed to a max. 60% luma value on the waveform, the white sheet of paper was around 65 on all three cameras.

For all cameras this resulted in a starting point as follows: shutter 1/25, ISO 800 (640 for the Panasonic S1) f4.0 All cams were set to record UHD 25fps, both the UMP’s were recording 4.6K which was scaled in post-production to UHD.

On both URSA’s (G1 with firmware 6.2 and G2 with firmware 6.4) we used the same Sigma 18-35mm f1.8 lens recording internal 4.6K BRAW constant bitrate 3:1. Both cams were allowed to warm up for 15mins, then a black shading (sensor calibration) was performed with the lens cap on.

The color checker in the image was used to set the base grade (using the respective tool in DaVinci Resolve) and deliberately, the same color grading settings were used on both cams to show differences (the G2 would have needed a higher tint value to begin with).

Test procedure
Starting with zero relative exposure, we set both URSAs to record, and successively reduced the exposure by increasing the shutter speed from 1/25th to 1/50th, then 1/100th , 1/200th then 1/500th and finally 1/1000th. Resulting in 0, 1, 2, 3, 4 1/3, 5 1/3 stops of underexposure. You might ask why 1/500th was used instead of 1/400th, well, the UMP’s don’t provide that shutter value – I should have used shutter angle instead. Next time.

Then in postproduction using DaVinci Resolve I pushed up the underexposure by the exact value in the camera RAW tab for both the URSA’s.

Results: URSA Mini Pro 4.6K G2

To give you an idea of how underexposed the images were, see below two images from the setup from the URSA G2 – the first one at zero settings, the second one at 5 1/3 stops underexposed.

URSA Mini Pro 4.6K G2 at the zero studio setting

URSA Mini Pro 4.6K at 5 1/3 stops underexposure

Now, let’s push the massively underexposed UMP G2 image back to zero.

URSA Mini Pro 4.6K G2 5 1/3 underexposed image pushed back to zero. Also, spatial and temporal noise reduction was applied (3 frames temporal, threshold 30 and spatial threshold 15). Unusable, as the vertical stripes cannot be removed by noise reduction.

Ok, it looks like we went too far.

URSA Mini Pro 4.6K G1 vs G2

Now, let’s look at the UMP G1 side by side to the new G2. In general, we can say that both cams start to show vertical stripes in the image, which become problematic at 3 stops of underexposure. The G1 starts to become pinkish, whereas the G2 becomes more and more greenish the more you underexpose.

URSA Mini Pro 4.6K G1 vs. G2 at 3 stops underexposure. At the cusp of being usable, unfortunately, because of the vertical stripes. The G2 behaves a tad better in my opinion but becomes greenish.

Not that it would make any sense, but here is another image from both cams at 4 1/3 stops of underexposure for your reference:

URSA Mini Pro G1 and G2 at 4 1/3 stops underexposure, pushed back to zero using the camera raw tab in DaVinci Resolve. Not usable.

Summary

The URSA Mini Pro 4.6K G2 shows a strong 12.6 stops dynamic range reading at a signal to noise ratio of 2 – a really good value! But what we learn from this test as well is that you have to be careful when shooting underexposed – the vertical stripes can easily ruin your image already at 3 stops under.

Are you shooting with any of the cameras mentioned in this article? What are your experiences? Let us know in the comments below.

36
Leave a reply

guest
1
1
- 2
Filter:
all
Sort by:
latest
 Naveed Parker-Nasir
Naveed Parker-Nasir
Member
January 29th, 2020

Great study and thanks for taking the time to do it. I’ve always felt the UMP didn’t have 15 usable stops of DR – my gut always said 12 max. I ignored my gut because I used to always confuse latitude with DR so thanks for emphasising the difference in this article. However last year I started a new workflow by converting my BRAW footage to Linear and then using colour space transform to convert my BRAW files to Arri Log C and then to Rec 709 – I would then applying luminance mapping and I’ve found on wide DR shots this seemed to improve DR performance and highlight roll-off substantially. I’ve had shoots where I’ve had to blow highlights and when using the Arri LOG C container in post I’ve managed to retain so much more information and range in the highlights as well as the rolloff than when using the standard BM 4.6K setting.

Crimson Son
Crimson Son
Guest
January 15th, 2020

Nice test. I know I have been critical in the past of the C5D’s methodology when evaluating camera picture quality. This definitely addresses my major concerns.
Charting of the noise factor really puts it in perspective.

Great job!

Viktor Yun
Guest
January 13th, 2020

Thank you for detailed lab tests! I have a plan to buy 4.6K G2 so your article helped me a lot. btw I have a question. What exactly the noise spectrum is, in 1st picture? I eager to know what the graph shows :)

Dean Butler, Alucinor Productions
Guest
January 11th, 2020

Just wanted to say I massively appreciate what you guys do!!! Thanks for sharing with us :-)

 Todd Peterson
Todd Peterson
Member
January 10th, 2020

Just before Christmas I got a G2. I went for it because it matches well with my primary cameras, Alexas, and has a lot in a sub-$6k package. I also got a BMPCC 4K with a speed booster. BMD was saying that this camera had 15 stops of DR. I know that’s mostly marketing, but in my own over/under exposure tests, I have been pretty disappointed in the G2. It’s only a tiny bit better than my BMPCC 4K. Yes it has internal ND, high framerates, etc., but I don’t feel like I’m getting 4-5x the value out of the G2 compared to the BMPCC 4K. The 4.6k resolution doesn’t really help me, and I don’t need lots of audio options and other things. Sadly, I think the G2 is going back. I had a lot of hopes for it.

Caine Mitchell
Caine Mitchell
Guest
January 10th, 2020
Reply to  Todd Peterson

Good to know your call on these two cams! Thanks for sharing Todd.

Dean Butler, Alucinor Productions
Guest
January 11th, 2020
Reply to  Todd Peterson

That’s interesting Todd. I shoot a lot of branded content/doco work. The onboard audio and nd’s has been awesome coming from Red. That said there’s a perceptible loss of range in the highlights and to my eye the noise is less “organic” looking.

 SMP Saint Marks Pictures LLC
SMP Saint Marks Pictures LLC
Member
January 11th, 2020
Reply to  Todd Peterson

Too bad. I find the G2 well worth the extra $. It guess it depends on the work you do. For my work (doc, narrative, branded ) internal ND’s are a deal breaker. But also the DR difference is quite noticeable. I also think the chip overall has a significanty better IQ. Also, operating hand held with the EVF and not having to “rig out” to make a camera functional is something I don’t miss.

 Саша Лубенский
Саша Лубенский
Member
January 13th, 2020
Reply to  Todd Peterson

Did a matching test just when it arrived – Alexa Mini, Epic MX, G2. G2 is pretty-much the same in terms of DR as Epic MX, ~11.5 usable stops. But when it comes to led lighting, G2 clips much faster than epic MX. G2 can’t reproduce basic color modes of Astera Titan Tube, making Magenta-Pink-Red all kinda the same color on european face. Personally i find a used Red much better companion to Alexa, otherwise I’d look for Sony Cinealta or older full-sized Alexa Classic. Varicams are superior in terms of matching color, but the form-factor makes them useless for me.

Dr.Janardon Gogoi
Dr.Janardon Gogoi
Guest
May 17th, 2020
Reply to  Todd Peterson

wow…I was searching for these words only. Sir, you made my day. But I felt the compact size and ND and the 120fps 4k would be a better option than my BMPCC4K. Now I guess it’s better to switch to BMPCC6K than putting way more money in the G2.

Daniel Rozsnyo
Daniel Rozsnyo
Guest
January 10th, 2020

Its the same sensor, only more of its output lanes were used, so one gets the double of frames per second in G2 over the G1 model. And that sensor still offers 2x more – maybe in G3 or G4, as well as it could do global shutter (at half the rolling shutter fps).

It is very easy for BMD to mislead all his customers into belief, that its a super cool new sensor. It is not. It is same one, and not it starts being a very aged one.

The difference in noise is made maybe by better calibration, difference in codecs, and a different sensor manufacturing lot. But not a different sensor model.

Guest
January 9th, 2020

The result is not very surprising. BM didn’t promise better dynamic range. I‘d like to mention that of course you can shoot at a shutter speed of 1/400. Therefore you push the shutter on the upper left of the screen. Then you tip on the main number on the lower left and a numeric keypad opens on the screen. Here you can choose the shutter speed you want.

Guest
January 11th, 2020
Reply to  Gunther Machu

Welcome Gunther.

Member
January 9th, 2020

Thank you Dr. Machu! The Dynamic Range Lab series is greatly appreciated. I hope the internal 12-bit Canon RAW of the 1D X mk III, 16-bit RAW of the FX9 and 12-bit ProRes RAW with the Nikon Z6 and Atomos Ninja V can please be tested in the future!

Jody Eldred
Jody Eldred
Guest
January 9th, 2020

Respectfully, as the cinematographer who facilitated the DR tests of the G1 at Dolby Labs in Hollywood with a world-renowned Ph.d engineer who determined the DR to be 15 stops (and Roy Wagner, ASC, who shot a feature with it– on which I served as the technology consultant– believes it to be in excess of 15), I choose to stick to their conclusions.

Johnnie Behiri
Admin
January 9th, 2020
Reply to  Jody Eldred

Hey Jody. Good to see you here.
Can you please share with us a link to this DR test and openly expose how it was done?

Thank you and see you in April? (NAB)

Johnnie

 SMP Saint Marks Pictures LLC
SMP Saint Marks Pictures LLC
Member
February 5th, 2020
Reply to  Jody Eldred

It wouldn’t matter if the tests were conducted by a kid in film school or Roger Deakins – to the readers at home, appeal to authority means little without showing us the data/footage.

 Łukasz Kalinowski
Łukasz Kalinowski
Member
January 9th, 2020

Hey! Guys good see that You testing it! i got G2, and to be honest i NEVER used a ProRes when i lern how to use BRAW. Its sooooo great solution and veru lightweigh for CPu and GPU, You can do a lot in post including extenssive color grading. Hilights recovery is also quite big what is missing in ProRes, and it really increasing dynamic range. I suggesst check it! Cheers! and keep rocking!

Caine Mitchell
Caine Mitchell
Guest
January 10th, 2020
Reply to  Gunther Machu

Love the sound of “runs buttery smooth” Gunther! Is that with Resolve or Premiere Pro native or w BRAW plugin or both Resolve and Premiere Pro?

Cheers guys.
Caine

Caine Mitchell
Caine Mitchell
Guest
January 10th, 2020
Reply to  Caine Mitchell

BTW thanks for sharing C5D :-)

 Ryan Hamblin
Member
January 9th, 2020

excited to see what the FX9 reports back with on dynamic range/latitude

Toni
Toni
Guest
January 9th, 2020

So, more or less except high frame rate G2 is the same as G1, right?! Good I didn’t buy “new sensor” joke. Waiting for something bigger from Ursa (like 6K full frame sensor with 15 stops dr).
?

 Łukasz Kalinowski
Łukasz Kalinowski
Member
January 9th, 2020
Reply to  Toni

G2 You have 300fps (HD crop) and You got 120 fps Full sensor (4,6k) Its Huge difrence and its worth! 120 fps and 300fps looks batter smooth and there is no artefacts of quality loss;D

Dean Butler, Alucinor Productions
Guest
January 11th, 2020
Reply to  Gunther Machu

Yep 120fps no sensor crop 4.6k is AWESOME!

 Саша Лубенский
Саша Лубенский
Member
January 13th, 2020

300fps looks as bad, as shooting in 2k on Red Epic MX. Massive quality loss, much softer image and the compression is waaaaay higher. Otherwise, i see no use for 300 fps – 120 is enought for most occasions, while when you do 300 – you always want it because you want 1000fps look, and it just doesn’t work. And, there’s one thing others usually don’t mention – G2 is heavier than G1.

 Kenny MacLaughlin
Kenny MacLaughlin
Member
January 9th, 2020

Great tests, thanks for doing this! I’ve wondered how it held up against the G1.

For the vertical lines – I own a G2 and I find that if you do a sensor calibration and let the camera warm up for a bit the lines become a bit less noticeable.

Unfortunately a lot of cameras have vertical lines like that. I did tests with a Red Gemini and it had a similar fixed noise pattern.

As for the rolling shutter, I hear it rates around 7ms.

1
1
- 2
Filter:
all
Sort by:
latest

Take part in the CineD community experience