Blackmagic Pocket Cinema Camera 4K

[Kisaha]

3 hours ago, drm said:

I updated all three of my cameras last night, using the 6.4 updater. The cameras now show a firmware version of 6.2.1, which is the last update on the BM site that mentions the P4K.

Don't they mention what's changed?

I hate those..Sony does them all the time also. Fuji is very respectful in that matter, very explainable.

[MeanRevert]

Pretty sure 6.4 was only for the Ursa Mini Pro. 

6.2.1 is the same as 6.2 except they fixed the problem they had with Windows users installing 6.2

[docmoore]

On 6/15/2019 at 8:47 AM, Jonesy Jones said:

Was there an official update that addressed highlight roll off? I hadn’t seen that. I’ve been waiting all this time for @Sage to help me with that. ?

Not from BM but a fairly interesting look at iso DR and overexposure with recovery of highlights.

 

 

[HockeyFan12]

On 6/14/2019 at 1:18 PM, majoraxis said:

From the Laforet Youtube video description: “What’s impressive about this package is the micro four thirds sensor. This sensor has an incredible roll off of highlights that I seldom see which really opens up your ability to capture an incredible dynamic range, from the deep shadows to extremely high highlights all the way to the sunlight”.

I wonder which Blackmagic Camera version update this was shot with to allow the sensor to display its "incredible roll off of highlights".  Were they shooting 6.1 with Cinema DNG? Were they shooting 6.3 or 6.4 and Blackmagic has silently rolled in better highlight response or was it there the whole time and the newest Resolve 16.4 beta or some future release beta unlocks the sensor "incredible roll off of highlight"potential with the way it process BRAW.

At 4:54 in this video, it illustrates and describes the highlight issue.

If Blackmagic has some how improved the highlight rolloff characteristics with pocket 4k generated video files,  it would be great to know what combination of Pocket 4k hardware version, Blackmagic Camera version update , recording format and recording format post processing to unleash this potential.

If nothing has changed on Blackmagic's side, maybe Laforet could share his recording technique to achieve incredible highlight rolloff with the Pocket 4k because this guy does beautiful work, so I am hoping he knows something I don't know and will be willing to share his knowledge with the community, or that he will tell Blackmagic and they can share it with the community.

Hopefully someone will clarify how this was made a reality.

I haven't used this camera yet (though the price is right), but I found this interesting:

At 400 ISO ("native"), you get four stops of highlight dynamic range. That's not bad, but it's not great. 

But you can imagine how someone shooting at 800 ISO would be happy with five stops over and someone shooting at 100 ISO would be very disappointed with two.

But the kids these days expose by histogram and not meter so who even knows anymore.

Besides that, rolloff has a subjective quality. The F5 has a pretty good over at 2000 ISO (six stops, I believe), but the original SLOG 2 implementation had chroma clipping and sharpening that look horrible.

I would try sticking to 800 ISO for challenging scenes if you can tolerate the noise level and see if that helps. I think the Alexa Mini has something like +7.8 stops at native 800 ISO (and even more at 1600 ISO) so that's insane.

[docmoore]

That is an old chart ... here is a screen capture of the updated DR with BRAW ....

 

ISO 640 may be a sweet spot for +6.1 stops ... and you have highlight recovery available if you do not blow the highlights.

[HockeyFan12]

47 minutes ago, docmoore said:

That is an old chart ... here is a screen capture of the updated DR with BRAW ....

 

ISO 640 may be a sweet spot for +6.1 stops ... and you have highlight recovery available if you do not blow the highlights.

Is that accurate? Where'd you find it?

That's a huge improvement. Really really good numbers, competitive (in practice, at least–shadows might be a little worse muddier) with the better cinema cameras available. 

Maybe that explains the difference. Those numbers indicate excellent highlight dynamic range. The early ones don't.

[tonysss]

I would like to see this chart for the original pocket. There must be lights in iso800 min. 8 stops.

[docmoore]

10 hours ago, HockeyFan12 said:

Is that accurate? Where'd you find it?

That's a huge improvement. Really really good numbers, competitive (in practice, at least–shadows might be a little worse muddier) with the better cinema cameras available. 

Maybe that explains the difference. Those numbers indicate excellent highlight dynamic range. The early ones don't.

Download PDF on this support page middle column fourth choice:

https://www.blackmagicdesign.com/support/

Page 42

 

Regards,

Bob

[webrunner5]

There is no way in hell that camera has 10.7 stops of DR @ 25,600 ISO. If it had 8 it would be amazing.

[HockeyFan12]

3 hours ago, docmoore said:

Download PDF on this support page middle column fourth choice:

https://www.blackmagicdesign.com/support/

Page 42

 

Regards,

Bob

Thanks. If we take 400 ISO as the BMPCC 4k base ISO, the revised numbers put base ISO over (highlight dynamic range) at +5.5. That's just behind the newer Reds at their base 800 ISO (numbers are unpublished, but similar to UMP 4.6k in practice), URSA Mini Pro 4.6k (+5.8, I think), etc. and just ahead of the C300 Mk I (+5.3). C300 Mk II is +6.3, guessing Varicam and Venice are around +6.5. Alexa is +7.4 to +7.8 depending on model... 

Or if you shoot at 640 ISO (+6.1) or 800 ISO (+6.5) then you've got highlight dynamic range competitive with big name cinema cameras. But you can also push other cameras a stop. Apparently Panavision rates the Red at 1600 ISO to get an extra stop. So then it's around +6.5, too, with more resolution.

Did braw also improve the highlights on the 4.6k? Couldn't find a manual for that.

1 hour ago, webrunner5 said:

There is no way in hell that camera has 10.7 stops of DR @ 25,600 ISO. If it had 8 it would be amazing.

I agree. The "under" part usually seems exaggerated, especially at high ISO. Canon claims something like 15 stops at 100,000 ISO and it's not even close. 

But the original specs on the BMPCC 4k indicate highlight dynamic range close to a 5D Mark III or a GH4 (+4 stops over 18% gray) and the new numbers are closer to high end cinema cameras. Maybe that explains the debate about highlight detail, and if the new numbers are accurate, it's pretty darned good...

Of course it also depends on how good the shadows look. I think the F5 and A7S both have +6 over at base, but their shadows are so bad in SLOG 2 you need to overexpose them by a stop. So it's less in practice. 

@docmoore Do you know if the revised over/under is only for braw or is that for ProRes, too?

[docmoore]

The ProAV Closer Look at Overexposure video ... linked above ... looks at both braw and Prores and they assume there is almost a stop of improvement

with the braw over the Prores ... but that was at +4 overexposure ... in the ability to claw back highlights.

In spite of the DR of the camera ... one really needs to expose so that nothing gets blown in the highlights ... when it goes it is ugly. Not the most gentle of

rolloffs. Zebras at 90 and watch false color for anything you want to be usable.

It is much better than first look would suggest ... but you have to rigorously control exposure.

 

[HockeyFan12]

1 hour ago, docmoore said:

The ProAV Closer Look at Overexposure video ... linked above ... looks at both braw and Prores and they assume there is almost a stop of improvement

with the braw over the Prores ... but that was at +4 overexposure ... in the ability to claw back highlights.

In spite of the DR of the camera ... one really needs to expose so that nothing gets blown in the highlights ... when it goes it is ugly. Not the most gentle of

rolloffs. Zebras at 90 and watch false color for anything you want to be usable.

It is much better than first look would suggest ... but you have to rigorously control exposure.

 

If I'm not mistaken that video is looking at Cinema DNG and ProRes, not at braw. They also show a graphic of the old chart that I posted, not the new one you found in the new manual.

The Cinema DNG highlight recovery looks okay, not great and not bad, but if braw is 1.5 stops better than that... well... that's really good. 

[docmoore]

You are correct ... the video is prior to the braw update. I assume that highlight recovery would be as effective as that of the Cinema DNG but that would

need to be demonstrated. Working with the raw files takes a bit of experience to find a decent balance that appears natural in high contrast landscapes.

Controlled lighting yields wonderful files ... great color and beautiful range of tones.

[HockeyFan12]

57 minutes ago, docmoore said:

You are correct ... the video is prior to the braw update. I assume that highlight recovery would be as effective as that of the Cinema DNG but that would

need to be demonstrated. Working with the raw files takes a bit of experience to find a decent balance that appears natural in high contrast landscapes.

Controlled lighting yields wonderful files ... great color and beautiful range of tones.

That makes sense. The CinemaDNG test is still a pretty good result now that I compare it with some of their other tests. 

I'm a little suspicious of ProAV's methodology:

Those cameras (C200, EVA1, FS7) are clipping at +4 over key on Caucasian skin. On average, Caucasian skin is about +1 stop over 18% gray. 4+1=5. So this indicates an over of <5 stops, somewhere around 4.5 stops it seems.

But the FS7 I believe is rated at +6 and the C200 is +6.3 and EVA should be around there, too. Suspicious. Highlight dynamic range is two stops short of expected. Or maybe that guy is really, really pale. How are they metering their scenes?

Compare with this test someone posted on Reduser:

http://www.reduser.net/forum/showthread.php?176662-FULL-FRAME-TEST-MONSTRO-ALEXA-LF-MAVO-VENICE&p=1857643&viewfull=1#post1857643

That indicates over values of:

>7 on Alexa (rated at 7.4-7.8)

>4 on Mavo LF

>6 on Venice 

>6 on C700 FF (rated at 6.3)

>5 on Monstro 

>5 on Ursa 4.6k (rated at 5.8 if I remember correctly)

That test seems much more accurate.

How is the performance for you at 800 ISO? This camera is getting more and more intriguing to me...

Edit: the more I look at ProAV's tests the more convinced I am they don't know what they're doing.

[docmoore]

I tend to shoot at 400/640 and then jump to 1250 to minimize noise .... so I have not explored 800 ... but I am very conservative in exposing

to protect highlights. But overall the camera delivers more than I expected ... but I may move to a S35 camera ... as I am used to the Z7 and Leica

SL ... as my two previous cameras. 

Mainly looking for a less build it yourself package ... and internal ND filters are high on my list ... as is TC in and out. I miss waveforms on the BM

although false color works ... a bit more finicky than a decent waveform in my hands.

[HockeyFan12]

1 hour ago, docmoore said:

I tend to shoot at 400/640 and then jump to 1250 to minimize noise .... so I have not explored 800 ... but I am very conservative in exposing

to protect highlights. But overall the camera delivers more than I expected ... but I may move to a S35 camera ... as I am used to the Z7 and Leica

SL ... as my two previous cameras. 

Mainly looking for a less build it yourself package ... and internal ND filters are high on my list ... as is TC in and out. I miss waveforms on the BM

although false color works ... a bit more finicky than a decent waveform in my hands.

Yeah, I've read you need to kit the camera out. I was looking into buying a Resolve license (15% off today at Rakuten, maybe I should just buy that) but also wanted a B camera for a web series I might shoot next year. (Talking heads mostly....) Was considering an SL3 or 90D when that comes out, but the BMPCC 4k would only be $1000 effectively after the Resolve license and I wouldn't be using it as an A camera.

I wrote this camera off before because I want more highlight detail (just personal preference), but +5.5 stops at base ISO is competitive with Red, etc. And 60p 4k raw is pretty awesome...

On the other hand, a dSLR would be smaller and more rugged, even if the BMPCC would be a ridiculously better A camera. Mediocrity may win again, probably going to end up with a 90D or something.

I found some strange discontinued Kenko variable ND that prevents color shift I think with a depolarization layer or something? Expensive (around $300) but originally around $700 and can be dialed in to as little as 1.3 stops. That said I am a huge fan of internal ND filters, too. 

Well this looks like an awesome camera that has only gotten better, but perhaps it is not for me after all.

[docmoore]

I would not dissuade anyone from the camera ... with a little planning it gives a drop dead gorgeous image ...

 

 

 

 

 

 

The quality of the Q 0 braw is worth the effort and file size. Down sizing and going from png snaps to compressed jpgs here do the files a disservice.

Regards,

Bob

 

 

[HockeyFan12]

44 minutes ago, docmoore said:

I would not dissuade anyone from the camera ... with a little planning it gives a drop dead gorgeous image ...

 

 

 

 

 

 

The quality of the Q 0 braw is worth the effort and file size. Down sizing and going from png snaps to compressed jpgs here do the files a disservice.

Regards,

Bob

 

 

Looks nice. Nice colors and tonality.

I'm taking this surprise sale on Resolve as a sign from the universe I don't need this camera, though. I have other bills and software I need to license this year that would agree. ? Will want them this year, won't need a B camera until next year is my guess. Hopeful this project gets off the ground, but not a lot of progress on it yet.

But braw sounds very impressive... the extra highlight detail and raw to SD cards seems to put this camera in a class of its own.

[docmoore]

Resolve is well worth the price ... indeed I bought it twice ... once from the Mac App store then again when I discovered I was hostage to Apple with regards to upgrades as they locked one out of all the beta improvements ...

And once purchased from BM it avails one of multiple upgrades without charge.

[CaptainHook]

There was no change in "highlights" to our log curves or otherwise. The previous manual dynamic range chart was incorrect, and was corrected.

There is also no inherent "highlight rolloff" in digital cameras from the sensor. They are all (mostly) linear capture devices. Same with Alexa, Red, etc. When they all clip in linear, they all clip suddenly/harshly. All of them.

 "Highlight rolloff" is really about total dynamic range, how you expose, and curves (which you can manipulate yourself).

Our log curves are optimized to store data in 10bit SDI and video files so the highlight roll off in our existing log curves is merely a function of optimising the dynamic range to fit in this data range with relation to the "native sensitivity" and desired ISO. The more dynamic range the sensor has the more highlights will roll off in the log curves, the same with Alexa/Red/etc. You want to use your own curves after in grading to create the roll off you want. Expose the sensor so middle grey is lower and leave more dynamic range in the highlights, and you can create a very nice gradual highlight rolloff in grading - with mostly any modern digital camera.

But it's pretty much all down to total dynamic range, exposure, and curves.