Blackmagic Pocket Cinema Camera has 10bit HDMI output

[Andrew Reid]

From the manual:

 

"Micro HDMI port of your Pocket Cinema Camera outputs 10-bit 4:2:2 HD video at 1920 x 1080 while recording"

 

I believe this is the first 10bit HDMI output I've ever seen on a consumer camera.

[Andrew Reid]

And still frame recording:

 

 

Time Lapse Interval

This setting allows you to record a still frame at the following intervals:

Frames: 2 - 10
Seconds: 1 - 10, 20, 30, 40, 50
Minutes: 1 - 10 

 

Should have a stills mode though, linked to shutter release. Still reading...

[peederj]

The important question on bit depth is are the bits actually used? You could have 12 or even 16 bit output and many of those bits will never carry relevant information due to the inherent DR of the sensor and the gamma being used. These are known as "marketing bits."

[Andrew Reid]

13 stops fits nicely into 12bit, converts nicely to 10bit.

 

Remind me, what is DR of C100 and the bit depth?

[peederj]

Rec 709, far from a linear mapping, offers 11 stops into 8 bits. Log gammas cram more stops into 8 bits.

 

I think it's good you don't have a C100+Ninja2 Andrew. You'd wonder why people were still discussing cameras and you'd go make movies instead. But I enjoy the blog and the banter, so keep struggling with half-measures.

[Andrew Reid]

Struggle. Haha. Half measures. Tee hee.

 

Pressing a record button on a DSLR isn't a struggle.

 

Getting 14bit film negatives off the card isn't a half measure.

 

Full frame sensor - no half measure there either!

 

I don't decry your personal reasons to shoot on what you shoot and neither should you mine, but looking at it objectively you have to admit you lost :)

 

The image is just flat out better!

[Andrew Reid]

Also if a $999 camera can make a profit and bundle 10bit HDMI and ProRes into the deal, why can't Canon do this at the $6000 mark? It's ridiculous.

 

10bit HDMI could be a real saviour of DSLRs. Panasonic should put on on the GH5 and feed uncompressed 10bit 4-2-2 LOG down it. They wouldn't need raw if they did that.

 

You forget that raw, aside from being an ultra creative workflow, is not all about 'raw' rather it is about bypassing the god awful H.264 mode of the 5D3.

[peederj]

Poorly done H.264/AVCHD 4:2:0 I will agree is regrettable, which is why I use the Ninja 2 with the C100 when I can. And I'd use it with the 5D3 too though the downsampling on that damages both the HDMI and the ML RAW similarly.

 

Absolute image quality is a nice-to-have but not a must-have compared to the practical realities of getting the shot in the first place. The C100 ergonomics and low light performance means an awful lot of shots are captured and captured better than they would with a high-ceremony camera. Add the Ninja 2 for direct-to-ProRes and you have a workflow that can actually be operated profitably and sworn by rather than abstractly theorized and sworn at. If you need IQ better than that you rent, full stop. The Alexa that nearly everyone rents is prized more for these practical factors than absolute image quality even though the image quality is outstanding too. Get comfy with a really well designed camera and it's hard to get motivated to complain about the failings of DSLRs and hacks, those are for fun not serious work.

 

I have the 5D3, with a 1000x card now, but I haven't tested the RAW yet. I use that for stills and as B-cam. I will test the RAW out when I get a multi-camera shoot that allows all that offloading to be justified in practice. That's a rare combination since multi-camera shoots are done for efficiently capturing ongoing events. Most even non-event shoots just aren't like that Andrew, you have moments to grab with a setting sun and lots of impatient people to juggle worrying about other things than IQ. It's not creating slideshow-like collages of nifty found scenes to demonstrate a camera output.

 

The 5D3's RAW is a good example of "marketing bits:" that camera offers about 11.5 stops (and that only at ISO100) and even in a standard contrast picture style will only use a little over 8 of those 14 bits you are fumbling with cards every ten minutes to preserve. The rest is noise or unused headroom. This is why a good 10-bit implementation for ML RAW would be costless if done well, and save you 30% wasted storage. MJPEG would be even better practically IMO.

 

I'm all for IQ and progress but I'm all against deceptive marketing and pound-foolishness. 4K 14-stop HFR RAW under $10K RTS will be nice in a couple years; until then there are <$10K RTS solutions that work and <$5K RTS half-measures for curious tinkerers. The extra $5,000 is a matter of whether you're actually doing this or just dipping your toe in.

[Andrew Reid]

Downsampling hardly hurts raw on the 5D Mark III. The image is very well sampled indeed and even performs on a chart. Did you see the joint Slashcam / EOSHD tests vs C100?

 

MJPEG is a terribly inefficient 8bit codec and very old, 1D C really suffers from it. Doesn't grade. Huge file sizes. ProRes is far superior.

 

"Most even non-event shoots just aren't like that Andrew"

 

Thanks for the real world lecture. Next time I am in my real world shooting at an abandoned Berlin sanatorium with actresses and a great script I will remember to cancel the whole thing because it doesn't fit your narrow minded view of a shoot FFS.

[HurtinMinorKey]

 

 

Remind me, what is DR of C100 and the bit depth?

 

12 stops of DR on the sensor.

[Andrew Reid]

Dear people of the DSLR community, including photographers. Herein lies an enlightening example of the problem with many pro video people as epitomised by Peejderj here. They are craftsman, even talented some of them, great understanding of the practicalities of shooting but they all have tunnel vision.

 

They put the practical realities of shooting AHEAD of everything else, including inspiration, intuition and spontaneity - those ingredients are critical to art-house filmmaking, narrative shorts and music videos.

 

The procedure is most important to them, and meeting a spec, delivering something quantifiable, by a set time.

 

And they are often incredibly arrogant. Only the shoots they have made a career doing are shoots, the rest is just pissing about, struggling away with workarounds, etc. Hey Peejderj... Upstream Color, GH2, enough said.

[Andrew Reid]

12 stops of DR on the sensor.

 

How much of it works its way into 8bit? Let's do the maths.

[peederj]

With Canon Log, all of it. 100%. With room to spare. There is a slight decrease in color fidelity due to the gamma vs. a linear mapping. But after I run it through Neat Video I can't see banding even on severe grades. Delivery codecs do far, far more damage than Canon Log into uncompressed 8 bit 4:2:2 via the HDMI out. Cinestyle HDMI out on the 5D3 v.1.2.1 is similar but not nearly as good in DR or resolution.

 

Anyway I've never known you to be arrogant Andrew. 

[AaronChicago]

Re: 10 bit. Doesn't the Pocket Camera record 10 bit 4:2:2 Pro Res to the SD card?

[Andrew Reid]

I will do a grading post on EOSHD soon which will show the difference between 14bit raw and 8bit codecs very clearly. Best thing to do is to stop talking about it and SHOW it.

[Andrew Reid]

 

Quick test. Massively blown window by 3ev.

 

Brought down 2.5 stops in post.

 

If you enjoy having your highlights look like the mush on the right, keep shooting 8bit / C100 / standard DSLR video!

[peederj]

Can't be the same exposure since the non-blown portions of the image aren't the same value.

[Andrew Reid]

Can't be the same exposure since the non-blown portions of the image aren't the same value.

 

LOL. It's the same exposure. What non-blown portions? The whole window including blinds were over exposed by 2 stops.

 

If you like I can upload the original H.264 with metadata for exposure, original raw DNG and even an original CR2.

[peederj]

To roughly test DR use a gray card in the frame and expose and post process for a defined value of that on your waveform/histogram. Then you will have a consistent baseline from which to measure latitude. You also need to note the gamma being used and normalize for that too. The original capture files and metadata would be useful of course. But this particular test isn't controlled and reproducible because sunlight is highly variable and may have changed drastically from one exposure to another merely by the presence of clouds.

 

What we need is a proper, scientifically controlled, reproducibly documented, fully independent test site for video as we have for stills (DPreview/DxO, which doesn't entirely meet my listed standards but gets closer than anything else). Then we can happily argue till we're blue in the face about subjective qualities...the objective ones will be established beyond argument.

[Andrew Reid]

Dude it was same shot on same camera with same lens and same exposure settings, same light same moment to within 5 seconds. You're an idiot.