4.6k beta tester footage

Here you go:

http://www.bmcuser.com/showthread.php?1 ... -Mini-4-6K


Sent from my iPhone using Tapatalk

And shot with rolling shutter, because GS has yet to employed. (source)

Very impressive readout speed of the sensor. Even at the end the scene with the cars - can't see the rolling shutter skew.

And shot with rolling shutter, because GS has yet to employed. (source)

Very impressive readout speed of the sensor. Even at the end the scene with the cars - can't see the rolling shutter skew.

yea - the RS really is very good. You CAN see if it if you whip pan or something shoots past quickly - but it is far less apparent than on the cinema cam. I predict people will use RS over GS for the majority of the time.

I'll repost my thoughts here:

It's a real spiritual successor to the bmcc.

It's funny how when trying to fit all the range into an end image, you often end up pushing it close to clipping anyway.

One of the things I am most excited about is the new film profile. It makes getting good and accurate images only couple of clicks away. Seriously, add a bit of contrast and you're there.

I love grading Alexa stuff, partly because of the image but mostly because of how their log image is to work with - I would go so far to say that the new film image is actually better to grade with than the Alexa's log image.

All this translates to faster resolve exports too- as you find yourself with half as many nodes.


Also, 800 is the nativa ASA - and so exposing for this gives you the best range - but this time, its even cleaner - less visible noise - so you have less of an excuse to ettr. Spent Sunday shooting just at 800 and exposing for it (no ETTR or ETTL) and it really shined. I refused to protect the highlights from clipping and it actually rarely happend (filming mid day in the UK winter) and when it did clip - it felt very natural.


Also the motion cadence is probably in Bolex territory - still evaluating this aspect though.

Playing with the SAT vs SAT and SAT vs LUM really shows how rich and deep the colour data is - even from 10bit ProRes HQ!


A big part of early testing with any camera is getting a feel for how best to use it - with the Cinema Cam, exposing for 400 gave you a clean image with good range - with the 4k Cam, protecting your highlights a bit helped with the range of that camera - with this one - I am pretty much decided that the key is to Expose for 800 (using false colour on the EVF) and even IF you somehow get clipping - I guarantee you will be fine with how it looks at that end. I wonder if this is the saturation point of increasing DR?

I havent done a lot of testing in low light yet - but I will say that my initial impression is that its probably up to a stop cleaner than the cinema cam (so 400 on the cinema cam is as clean as 800 on the 4.6k - 800 on the cinema is as clean as 1600 on the 4.6k). Does this mean you can push it digitally in post to higher values? sure you can - but I think the limits of digital gain will kick in on any camera, even if its cleaner. If you liked the natural grain style of the cinema camera, this one is the same. Its not nasty noise - its quite pleasant.

This was shot at 1600 and had a very very simple grade applied. It was not pushed. Remember also that its entirely lit by fire - which some of you may know gives a very unforgiving colour temperature for noise. The room was very very dark.





A single cheap Fluro Source (hardly a good light source) - ProRes HQ - 2 nodes.




No clipping in this shot either - shot at 1080p ProRes 422 (not even HQ) at ISO 800 -

NO noise reduction applied, it was exposed in the middle - some sharpening applied.

You can see some CA from the Canon 24-105 lens I was using (at f5.6 if I recall correctly)

So this has been downscaled from the full 4.6k sensor resolution to 1080p In camera.




If you forgive the bad white balancing here - have a look at the difference between Raw Lossless (10.5MB per frame) compared to 3:1 Lossy (4MB per frame)

The fabric backdrop has that pattern on it, its not moire.

I didnt label which was which before - and I cannot remember now.

I remember pixel peeping at 300% in the shadows and could see a slight difference in the 3:1, less detail - but only but a tiny amount, and it was so close to 0% anyway then its hardly much a of a loss.

Any questions?

(preferably asking opinions rather than tech specs etc - BM are the best people to ask about that)

Mind blown - can't wait to give that baby a testride.

Mind blown - can't wait to give that baby a testride.

I think we can all agree we'd love to see Frank take it for that test drive. Make it happen Blackmagic Design!

Here you go:

http://www.bmcuser.com/showthread.php?1 ... -Mini-4-6K


Sent from my iPhone using Tapatalk

some of the clips looked REALLY good, specifically the black vest/white shirt combo. (it was party due to good, overcast lighting, though, i'm sure.)

...and then, some of the shots looks very 'video' to me. is this just the result of of ProRes? i'm used to shooting everything in Raw.

Now, i'm not saying the DR isn't there, not at all, but it didn't really look that "cinematic" to me. does that make sense? I understand the 'slight color grading' and the 'on occasion herky-jerky handheld shots' but i didn't really see a filmic vibe going on. bleh?

Now, i'm not saying the DR isn't there, not at all, but it didn't really look that "cinematic" to me. does that make sense? I understand the 'slight color grading' and the 'on occasion herky-jerky handheld shots' but i didn't really see a filmic vibe going on. bleh?

Same. Likely down to the "grading" though. I'll wait for straight from camera footage before judging too harshly.

yea - give it time - its a bit of a change from spending most of my time grading for the cinema and 4k cam - suddenly I have a lot more range to work with, and in all honesty - sometimes Its hard to decide how to use the full range of the camera.

I have shot about 3 full days externally on the camera (and a whole load of smaller test shoots, comparisons etc) so I have a lot of footage to go through and present. I am sure a more cinematic/filmic image is possible - certain of it - but when youre just getting to grips with a new camera, one tends to try and find the neutral point first - then be more experimental with the grade.

Know what I mean?

Thanks for testing Tom, very exciting times. Looking forward seeing more of this.

I may sound stupid, but could this film profile also be implemented in the 2.5k?

Thanks for testing Tom, very exciting times. Looking forward seeing more of this.

I may sound stupid, but could this film profile also be implemented in the 2.5k?

I have no idea - I don't know how much of it is because of the superior sensor in the Mini and how much is just a different approach that they have taken.

I guess there is a legacy issue to consider - if you go and change how a camera shoots then it renders any pre-made LUTS for the camera useless, as well as if anyone wants to go back and re-shoot to match.

just a quick frame grab - shot at 400 - ProRes 422 1080p - no sharpening or NR - just a quick balance (left a bit warm for style) and then a bunch of Impulz luts applied.

That is the most delightful test footage I've ever seen. I love that song, too. Roman did a great job.

This preview does two things: 1. Makes the wait harder, 2. Makes the wait even more worth it. Now we know we're getting closer to release as test footage is releasing and NDA's are being slowly lifted. I sincerely can not wait to get mine and see what others are doing with theirs. Even with a simple one node grade this camera is certainly able to capture some fine images. A lot if that test video looked like a modern advertisement for...something fashionable. I have not been this excited since the announcement of the original Cinema Camera. As soon as I can think of some questions...oh yeah: power draw. What batteries are you using and how long are they lasting?

Some shots with a grade I have been working on for the mini


Again - all these shot 1080p ProRes 422








yet again, no clipping in any of these

Good to hear there's development. I'm not so sure how I feel about the new log mode being contrasty. I personally like the flat output and wasn't a huge fan of the contrasty output of the 4k sensor. I really like the pocket and cinema cam...I was hoping it would be nice and flat but I guess that's just me.

Good to hear there's development. I'm not so sure how I feel about the new log mode being contrasty. I personally like the flat output and wasn't a huge fan of the contrasty output of the 4k sensor. I really like the pocket and cinema cam...I was hoping it would be nice and flat but I guess that's just me.

I wouldn't say its contrasty - its more that the HUE is correct and the Saturation is richer. It is closer to the Alexa's LogC - brighter mid tones.

If you added some contrast in both the LUMA and Chroma then you would arrive at a good graded image. No need for secondary corrections or adjustments in order to get an accurate image.

It's not like the 4k FILM profile at all.

Good to hear there's development. I'm not so sure how I feel about the new log mode being contrasty. I personally like the flat output and wasn't a huge fan of the contrasty output of the 4k sensor. I really like the pocket and cinema cam...I was hoping it would be nice and flat but I guess that's just me.

I wouldn't say its contrasty - its more that the HUE is correct and the Saturation is richer. It is closer to the Alexa's LogC - brighter mid tones.

If you added some contrast in both the LUMA and Chroma then you would arrive at a good graded image. No need for secondary corrections or adjustments in order to get an accurate image.

It's not like the 4k FILM profile at all.

Ah ok I got ya...that's good news.

Some shots with a grade I have been working on for the mini


Again - all these shot 1080p ProRes 422


yet again, no clipping in any of these

I really, really like how life-like those shots looks - especially the first one with the trees.
I do not care so much about blockbuster looks, but I am all for those looks looking like the real thing. I think the extra 2 stops will improve that very much in the highlights.

On some of the samples I see halation around over-exposed highlights. While that's generally a desired look, is that an artifact of the sensor or the lenses you're using?

Also, how is the highlight roll-off when you look at it without grading? Are you using the 'soft clip' (or whatever it's called) in Resolve?

On some of the samples I see halation around over-exposed highlights. While that's generally a desired look, is that an artifact of the sensor or the lenses you're using?

Also, how is the highlight roll-off when you look at it without grading? Are you using the 'soft clip' (or whatever it's called) in Resolve?

If you're referring to my samples - only the fire shot has any clipped sections. Most of the shots (apart from the fruit) have been shot on either the Samyang 35mm 1.4 (which can be soft) and the Canon 24-105L - which is also not a very sharp lens. So perhaps its just the characteristics of the lenses?

Hopefully resolve will get CA correction someday. Adobe Camera Raw does a brilliant job at this.

Did a quick grade and edit of this this evening for you guys - I don't recall any clipping in any of these shots either (which to be fair I don't expect from such flat light)







Shot on the Blackmagic Design Ursa Mini 4.6k - EF
All in ProRes 422 1080p - some shots in 60fps, some in 120fps
The intention was to create a natural and realistic grade - the natural light was very bleak and foggy.
This grade was achieved in just 3 nodes.
Everything filmed with the Samyang 35mm 1.4 Lens.

Did a quick grade and edit of this this evening for you guys - I don't recall any clipping in any of these shots either (which to be fair I don't expect from such flat light)







Shot on the Blackmagic Design Ursa Mini 4.6k - EF
All in ProRes 422 1080p - some shots in 60fps, some in 120fps
The intention was to create a natural and realistic grade - the natural light was very bleak and foggy.
This grade was achieved in just 3 nodes.
Everything filmed with the Samyang 35mm 1.4 Lens.

Wow dude, this is brilliant, 1080p is realy sharp. Any chance you can also find time to upload a 4.6k footage onto youtube?
Thanks for the test also

[CaptainHook]

The log curve for this sensor is flatter than any of our previous cameras, because of the extra dynamic range. But we've tried to make it easier to grade than previously at the same time.

Bleak and foggy conditions make high DR shine! Thanks for sharing Tom, looks great.

The log curve for this sensor is flatter than any of our previous cameras, because of the extra dynamic range. But we've tried to make it easier to grade than previously at the same time.

I don't get one till April, May time but thats ok. love that the testers are putting stuff out now.

On some of the samples I see halation around over-exposed highlights. While that's generally a desired look, is that an artifact of the sensor or the lenses you're using?

Also, how is the highlight roll-off when you look at it without grading? Are you using the 'soft clip' (or whatever it's called) in Resolve?

Both the 4.6k test footage on the BMD site and the sample Ursa mini 4.6k footage on youtube by Roman have varying levels of diffusion on the lenses making the highlights bloom like that.

On some of the samples I see halation around over-exposed highlights. While that's generally a desired look, is that an artifact of the sensor or the lenses you're using?

Also, how is the highlight roll-off when you look at it without grading? Are you using the 'soft clip' (or whatever it's called) in Resolve?

Both the 4.6k test footage on the BMD site and the sample Ursa mini 4.6k footage on youtube by Roman have varying levels of diffusion on the lenses making the highlights bloom like that.

I should point out - none of my shots have diffusion applied. No filtering of any kind in fact.

Instead Tom, but your somewhat "soft" lenses add diffusion of their own, which gives your shots a nice look. Thanks for sharing your outstanding shots and experiences with the camera. How is the audio?

Instead Tom, but your somewhat "soft" lenses add diffusion of their own, which gives your shots a nice look. Thanks for sharing your outstanding shots and experiences with the camera. How is the audio?

I'd think the audio would be the same with the mini.

I would tend to agree. Note, how is your experience with the audio, have you tested it with a good mic connected directly vs using your SD MixPre With the camera? Thanks

I would tend to agree. Note, how is your experience with the audio, have you tested it with a good mic connected directly vs using your SD MixPre With the camera? Thanks

Yes I'll link you. Though interestingly there might be truth if they differ such as if the 4.6k has gyro or not. I'll message when I get back.

Does the 4.6k have a gyro?! :p

I'm sold. Sadly, putting my money into an iMac 5K first. But, then I'm saving up to upgrade to the URSA Mini 4.6K! Or, I rob a bank. Either way, the imagery looks amazing. I really want to play around with it. Thanks for sharing guys.

Great job Beta Testers.

I think I found the camera I would like to shoot my narrative ski films. As well, every other movie idea in my head.

OK folks, Here is another bit of demo footage I shot very early on.

Shot in ProRes 422 again, a few shots in UHD, mostly in 1080p. Some Crop sensor mode 120fps, some full sensor 60fps.

In some of the running shots there is some clipping here and there - but not much. Again - all Rolling shutter here.





One thing I did learn from this shoot - Lens IS on the Canon 24-105 is just not good enough for any HFR filming. It reacts too slowly and introduces a robotic jerk - something which I never saw on 24/25fps footage previously (except in panning shots)

Yes, the Canon (and most other DSLRs is designed for still shooting, or very simple video shooting on their cameras. Put the lens on a camera like the BM Mini 4.6 and pushmthemfrsme rate, you will see the physical limitations of the lens IS system.

I would tend to agree. Note, how is your experience with the audio, have you tested it with a good mic connected directly vs using your SD MixPre With the camera? Thanks

https://www.dropbox.com/sh/j6nnjayc04tu ... 8NhDa?dl=0

Thanks Note, will get this downloaded.
Cheers

Thanks Note, will get this downloaded.
Cheers

No problem. Also I don't have the Mini anymore (not until the 4.6k gets here :p).

Oh, sorry to hear that, you were having so much fun! Cheers

I'm seeing quite a bit of FPN in the lowlight shot of the woman at 02:49 in video that Roman posted. I don't mean to just point out negatives. But that's a bit discouraging.
When there's enough light hitting the sensor it looks great, as is the case with most cameras, I suppose. And I know this is early footage, but based on that it doesn't look like pushing it past 800 or digging into the shadows are going to be options.

I'm seeing quite a bit of FPN in the lowlight shot of the woman at 02:49 in video that Roman posted. I don't mean to just point out negatives. But that's a bit discouraging.
When there's enough light hitting the sensor it looks great, as is the case with most cameras, I suppose. And I know this is early footage, but based on that it doesn't look like pushing it past 800 or digging into the shadows are going to be options.

I'm not seeing the same FPN that you're seeing apparently.

Are you thinking of the compression artifacts from YouTube?

I'm seeing quite a bit of FPN in the lowlight shot of the woman at 02:49 in video that Roman posted.

Where? I see moving noise but not Fixed Pattern Noise.
And only when blowing up this video on a 4K screen.

It's amazing how many people think that YouTube videos are an even adequate representation of a camera's image quality.

Look at the woman's face in that shot. And also the blue sky behind her.

And, no, Rakesh, I don't think that web videos are representative of image quality. But if you can see issues in spite of compression, they will be there in the original footage as well. I would guess that the web compression actually makes it less obvious.

I'm always happy to be wrong. But from here it looks like FPN.

Look at the woman's face in that shot. And also the blue sky behind her.

And, no, Rakesh, I don't think that web videos are representative of image quality. But if you can see issues in spite of compression, they will be there in the original footage as well. I would guess that the web compression actually makes it less obvious.

I'm always happy to be wrong. But from here it looks like FPN.

Yes it looks like it to me as well.

Web compression will usually make problems with the footage less obvious by introducing worse ones.

Looking forward to seeing some ungraded and graded (without the application of a LUT) raw footage.

Hey all, first post here after reading and following a number of very useful threads for a while! Please don't be too hard on me haha. I may be still a university student, though I do have a bit of a background on the technical side of things with my physics double major (as well as a lot of detailed reading and investigation into the such topics on my own), so I wanted to share my thoughts on this.

With the compression levels common on Youtube and Vimeo, it is difficult to confirm the true source of the slight artifacts seen in a few other scenes from this video and others, so we can't presume anything there, and they aren't particularly noticeable to the casual viewer unless one is looking for them. However, in the one shot Paris Remillard pointed out, there is some rather obvious chroma (and likely luma) noise visible on the face and in much of the sky (likely in the darker regions too but less visible due to the crushed blacks). However, as Paul Knapp pointed out, this is not FPN by definition as it is randomly distributed both spatially and temporally, characteristic of shot, or photon noise, as opposed to FPN found in a set grid due to small but consistent differences in pixel sensitivity made visible due to the low-light (low SNR) conditions.

This distinction may seem academic, but is in fact quite critical and relevant to your discussion. FPN can be corrected or virtually eliminated by better sensor design, QC, and in firmware as BMC claims to have done, as well as with certain techniques (eg dark frame subtraction) in post. This is similar to read noise, random errors in measuring the light that hit the sensor that limits its shadow DR (since with less light/"signal" coming in, random electrical noise and read inaccuracies are relatively larger), in that a perfect sensor could, theoretically have zero. However, shot noise is a much harder to avoid since it runs up against a fundamental limitation of physics: think of each photon as the smallest packet of light that a sensor can measure, and since at the nano-scale these are randomly distributed (Poisson distribution, technically, but similar to a normal bell curve) an individual pixel "bucket" might catch a few more or less than its neighbors, or than it did in the frame before or after the current one. Again, this only becomes a problem at low signal (light) levels for the same reason we have log gamma: to the eye, the difference between 112849 photons and 112857 photos is vastly less than between 20 photons and 30 photons.

This is a fundamental physical property of light and can only be changed by measuring more of it at once, to get an average closer to the "true" mean that's less noisy (kind of like polling more voters to get a better handle on the outcome of an election). We can do this three ways: by gathering a wider chunk of light (aperture), by measuring for longer (shutter), and (crucially) by converting a higher percentage of photons that hit the sensor into electrons that we measure, which is called quantum efficiency. This is already pretty high for most modern sensor materials, close to 50%, but the big difference comes from how much of the sensor's area is devoted to gathering light. In a FSI sensor, the sensor's readout electronics are in front of the actual detectors, like in the human eye, but in a BSI sensor, the opposite is true (like the cephopod eye) allowing more of the photons to make it to each detector (pixels). The BIG deal here is that the global shutter design of all the recent BM sensors takes up a lot of this space (whether you are using it or not), meaning that the sensor gathers fewer electrons per amount of light and thus a higher shot noise relative to RS cams like the C300 and FS7.

So, the upshot is that at some point, with any camera, you're going to hit the shot noise floor and there aren't a lot of easy ways around that, particularly without massive R&D budgets. However, some of the noise *could* be coming from the sensor's analog amplifiers (or lack thereof) that are locked to a much lower level than those on the FS7 or C300 can go given the camera's single analog ISO (nominally 800), and thus without the ability to amplify the original signal (shot noise and all) up to a higher level relative to the fixed amount of noise inserted at the readout (A/D conversion) stage, your going to hit that limit first anyway. Impossible to really tell at this point, nor does it matter much for this cam if the sensor hardware is already fixed (but can be more easily fixed on the hardware side in the future with either better A/D converters or (variable) analog preamps, and possibly even tweaked a *little* with low level firmware. For either one, the solution is the same for the user without analog ISO: larger physical aperture lenses, longer shutter speeds, or NR in post.