Blackmagic Pocket Cinema Camera 4K

[Savannah Miller]

1 hour ago, webrunner5 said:

The original Arri Alexa did not do Raw other than to an external recorder. And at the time of introduction Arri claimed the following. As you can see they were Only claiming 14 stops.  You have to figure the Alexa is "Only" outputting Arri RAW 2.8K in 12 bit. So it is not going to be super high.

Long story short a lot of cameras really don't put out as much DR as you would think. But I want to correct my above statement about the DR of the Canon 3D mk III. I said 9.5, I meant 10.5 maybe 11.

 

ALEXA Classic EV Features

BEST OVERALL IMAGE QUALITY

Film-like, organic look

High Dynamic Range

14+ stops exposure latitude over the entire EI range as measured with the ARRI Dynamic Range Test Chart

future proof for High Dynamic Range (HDR) displays

16 bit in-camera processing

 

But there is this article that was interesting at the time.

https://nofilmschool.com/2014/07/sony-a7s-dynamic-range-arri-alexa-amira

And this older one.

https://nofilmschool.com/2013/03/blackmagic-cinema-camera-red-epic-arri-alexa-raw-test-part-3-overexposure

 

Resolution doesn't affect dynamic range.  Resolution only increases detail when you downscale so when maybe you get 1/2 stop more "resolved" details in the shadows vs. a lower resolution image.  The Alexa and BM cameras use dual sensor readout to get their high dynamic range perfomance.  RED builds high dynamic range sensors with high resolution, but because the photo-site size is smaller, they've always danced around the Alexa in terms of dynamic range.  Dragon was pretty close to Alexa DR but Helium was less.  Maybe the weapon processing of dragon improved things, but it still wasn't noticeably more if than the Alexa unless you lift up the shadows and denoise.  Gemini and Monstro are large photosite sensors.  Gemini is only 5K, but Monstro is 8K large format.  Those sensors are now probably more DR than the Arri Alexa, but they're still not a lot more.

[squig]

2 hours ago, John Brawley said:

Actually RED initially were claiming 20 stops when they were first releasing images. Anyone else remember this ?

www.thephoblographer.com/2013/01/01/reds-dragon-sensor-lives-up-to-hype-of-20-stops-of-dynamic-range/amp/

and sure you can count 20 stops in that image. 

But we all know they don’t really have 20 stops. 

So anytime someone makes a claim about stops (including Blackmagic) I always judge by actually shooting something, preferably alongside another known like an Alexa. 

JB

The backyard test (paling fence in full shade, blue sky, fluffy clouds) sorts the actual from the aspirational dynamic range. BM marketed the BMMCC as 13 stops. I tested the BMMCC raw against the 5D Mk3 raw (11.7 stops) and the 5D had more latitude in the grade.

Good write up on dynamic range vs latitude: http://www.xdcam-user.com/2013/11/whats-the-difference-between-latitude-and-dynamic-range/

[Savannah Miller]

5 hours ago, squig said:

The backyard test (paling fence in full shade, blue sky, fluffy clouds) sorts the actual from the aspirational dynamic range. BM marketed the BMMCC as 13 stops. I tested the BMMCC raw against the 5D Mk3 raw (11.7 stops) and the 5D had more latitude in the grade.

Good write up on dynamic range vs latitude: http://www.xdcam-user.com/2013/11/whats-the-difference-between-latitude-and-dynamic-range/

Latitude might be due to your skill and using Adobe camera raw.  BMCC has more latitude in the grade than the 5d mark III, you just have to know how to work it.  Adobe camera raw is a bit weird and not ideal for the camera.

[squig]

2 hours ago, Savannah Miller said:

Latitude might be due to your skill and using Adobe camera raw.  BMCC has more latitude in the grade than the 5d mark III, you just have to know how to work it.  Adobe camera raw is a bit weird and not ideal for the camera.

I'm talking about the Micro Cinema Camera, not the BMCC. I graded both in Resolve.

[John Brawley]

8 hours ago, squig said:

The backyard test (paling fence in full shade, blue sky, fluffy clouds) sorts the actual from the aspirational dynamic range. BM marketed the BMMCC as 13 stops. I tested the BMMCC raw against the 5D Mk3 raw (11.7 stops) and the 5D had more latitude in the grade.

Good write up on dynamic range vs latitude: http://www.xdcam-user.com/2013/11/whats-the-difference-between-latitude-and-dynamic-range/

How do you measure stops on a "backyard" test ?

Again the only meaningful way to measure in my view is to shoot side by side in the same situation.  Then you can say categorically one is better than the other.

Like when I comapred an Alexa to an Ursa Mini 4.6K here

https://johnbrawley.wordpress.com/2016/03/23/ursa-mini-4-6-and-alexa-side-by-side/

I think all camera manufacturer claims are kind of meaningless.  They're almost ALWAYS derived by calculating a theoretical dynamic range from the sensor SNR and not from actually photographing anything. They don't sit their with charts in a lab.  The do it with maths from a spec sheet.   Arri are the only company that have actively UNDERSTATED their claims of DR. 

Most accept that an Alexa is a stop and half better than what Arri claim.

Arri also make their own DR chart which I've never seen anyone use publicly..

https://www.arri.com/arriajax?mod=productList&product=263

The actual photographable DR is pretty much impossible to measure anyway because none of us can agree on how much noise is too much.  What you think is acceptable I may not. So who is right ?

Noise is subjective. Everyone feels differently about how much is acceptable.  That's why oyu can post a picture of a 20 wedge chart and say hey it's got 20 stops !

And by the way that article trying to make use of the word "latitude" is also highly subjective.  It's some individual person's SUBJECTIVE view about what's USEABLE DR. So what is useable DR ?  What does that mean in a SCIENTIFIC repeatable way ?   Latitude is another meaningless phrase like "useable dynamic range"

What does useable mean to that one individual ?

In what scenarios does that change ?  Low light / low contrast over daylight high contrast ?

The only time I see the phrase "useable dynamic range" being useful is when you're in log encoded video and you're trying top re-position a tone that's been recorded near clipping and when you try that it can look pallid and weak.

So you have a skin tone that isn't technically clipped but when you put it from 90% and put it at 30% it looks crappy and weird.

This is not a limitation of dynamic range per se, but the shortcoming of LOG encoding video curve being used and assigning less "bits" to that part of the curve.  It's a form of compression.  

Arri do 16 bit linear-->12 bit log at RAW.  So do Blackmagic when shooting DNG.  Sony RAW is also 16 bit LIN @ 3:1 (on an F55 anyway)

JB

 

 

 

 

 

 

 

 

 

[Anaconda_]

For anyone wondering about the Samsung T5 drives, I just picked a 500gb one up on sale. Here's a speed test connected via USB 3.0 to my MacBook. Only posting because I hadn't seen it anywhere here, and thought it might be useful

[webrunner5]

On 6/25/2018 at 5:50 PM, Savannah Miller said:

Those 18 stop RED claims are ridiculous.  Download the R3D's of the xyla chart and try them out for yourself in davinci resolve.  It's VERY hard to produce the dynamic range numbers they are claiming.  Those xyla chart images are misleading as they use frame averaging of multiple frames to remove noise and they use multiple ISO settings within the R3D and composite the results together.  That's not how you use a xyla chart.

 

The sensor in that video that's 18 stops is the Dragon 8K VV sensor which was never released.  Now it's the monstro sensor which supposedly is a little better.

 

In order to reveal the dynamic range numbers phil is getting, I have to set the ISO to the highest settings and crank it up further just to dig into the shadows. Even then I still can't get as much information as he is clearly getting so I have no idea what he's doing. If the camera captures 18 stops of DR, it's very hard to reveal it all and impossible to use.  

 

For a long time people on there was a long thread where asked one simple question:  If Red Dragon and Alexa captured xyla images and were processed in exactly the same method would the Alexa show the same or better dynamic range than the dragon?  Everyone was dancing around the answer when the main tester who was comparing both cameras eventually broke down and said that he believed the Alexa would show more dynamic range.  Likewise, there are tons of tests showing that without digging up lots of noise, the Alexa still does have more usable DR than helium and dragon cameras.

 

There's even old RED images showing alexa and RED MX xyla chart images side by side with a minimal DR difference.  But we all know that isn't remotely true.

I had forgot about this Tom Antos video on how he tests DR. He uses the Waveform Monitor. Seems to work pretty reliable as far as I am concerned. And you can make the same test yourself with other cameras pretty cheaply.

 

[squig]

8 hours ago, John Brawley said:

How do you measure stops on a "backyard" test ?

I don't. I look at real stuff like highlight roll-off and shadow noise.

8 hours ago, John Brawley said:

Again the only meaningful way to measure in my view is to shoot side by side in the same situation.  Then you can say categorically one is better than the other.

Which is exactly what I did/do. Not just the backyard: tungstens, low light, etc.

8 hours ago, John Brawley said:

I think all camera manufacturer claims are kind of meaningless.  They're almost ALWAYS derived by calculating a theoretical dynamic range from the sensor SNR and not from actually photographing anything. They don't sit their with charts in a lab.  The do it with maths from a spec sheet.   Arri are the only company that have actively UNDERSTATED their claims of DR.

Mos def.

8 hours ago, John Brawley said:

The actual photographable DR is pretty much impossible to measure anyway because none of us can agree on how much noise is too much.  What you think is acceptable I may not. So who is right ?

Noise is subjective. Everyone feels differently about how much is acceptable.  That's why oyu can post a picture of a 20 wedge chart and say hey it's got 20 stops !

Exactly. And different cameras have different noise characteristics like fixed pattern, blotchy, chroma... some noise is better/more tolerable than the rest.

8 hours ago, John Brawley said:

And by the way that article trying to make use of the word "latitude" is also highly subjective.  It's some individual person's SUBJECTIVE view about what's USEABLE DR. So what is useable DR ?  What does that mean in a SCIENTIFIC repeatable way ?   Latitude is another meaningless phrase like "useable dynamic range"

What does useable mean to that one individual ?

In what scenarios does that change ?  Low light / low contrast over daylight high contrast ?

Hence the need to do your own tests rather than just buying on spec.

8 hours ago, John Brawley said:

The only time I see the phrase "useable dynamic range" being useful is when you're in log encoded video and you're trying top re-position a tone that's been recorded near clipping and when you try that it can look pallid and weak.

So you have a skin tone that isn't technically clipped but when you put it from 90% and put it at 30% it looks crappy and weird.

Pasty?

[kye]

9 hours ago, John Brawley said:

Noise is subjective. Everyone feels differently about how much is acceptable.  That's why oyu can post a picture of a 20 wedge chart and say hey it's got 20 stops !

(Apologies if you already know these things) In audio they have a number of measurements related to relative volume levels or Signal-to-Noise ratios, maybe a similar mechanism would be good.  No idea how to get an industry to adopt a standard term though!

55 minutes ago, squig said:

Exactly. And different cameras have different noise characteristics like fixed pattern, blotchy, chroma... some noise is better/more tolerable than the rest.

In line with my above comment, this is also a problem in audio, and unfortunately they haven't nailed it yet.  Different types of distortion have different aesthetics but will measure the same in absolute levels.  I suspect that things like Motion Cadence and Colour Science will be quantifiable, we just haven't worked it out yet.

[Savannah Miller]

13 hours ago, squig said:

I'm talking about the Micro Cinema Camera, not the BMCC. I graded both in Resolve.

BMCC, Micro, BMPCC are basically the same sensor and color science.  The pocket and micro is a smaller cut, but it's pretty much identical.  RED does not lie about the SNR of their sensors, but they do claim higher than theoretical dynamic range numbers.

In resolve it's a bit unfair too as the color transforms from the Magic Lantern Raw are probably not accurate.  I'm sure the Blackmagic has not only more DR but more latitude as wel.  It just comes down to the skill of the colorist.

 

The 5D sensor likely has less noise though so you can lift the image more without seeing problems, giving the appearance of more latitude.

[webrunner5]

To me noise doesn't make much difference when you are doing a DR test. To me the last last stop on the graph is the one that you can barely see that is the end of being usable, the next one is in a sense is nearly invisible, so does not count . So noisy or not you could not see it anyways. Now I know you can't have crazy noise but real film was noisy as heck, 8mm was a good example. And people even add it, so noise is not the end of the world.

[squig]

28 minutes ago, Savannah Miller said:

BMCC, Micro, BMPCC are basically the same sensor and color science.  The pocket and micro is a smaller cut, but it's pretty much identical

Too many bloody acronyms. It's interesting what you said about the BMCC having more latitude in the grade than the 5D Mk3 raw. Did you ETTR with the 5D?

I found the BMMCC to be a bit too noisy for my liking, and it didn't like tungstens: yellow turned to orange. I expect the Pocket 4k will be much improved in both areas.

[jonpais]

16 minutes ago, Savannah Miller said:

The 5D sensor likely has less noise though so you can lift the image more without seeing problems, giving the appearance of more latitude.

But I thought being able to lift shadows in post with less noise was the definition of more latitude... ?

[Savannah Miller]

27 minutes ago, jonpais said:

But I thought being able to lift shadows in post with less noise was the definition of more latitude... ?

Well yes but that's only if you're not clipping.  If you take an scene that has less than 11.5 stops of DR then yes, it would appear that he 5D has more latitude.  But in a high contrast scenario it's pretty clear that the pocket has both more DR and more latitude.

[mercer]

Well, I think it’s safe to say that the 5D3 with ML Raw has close to 12 stops and the BMPCC has close to 13... inconsequential.

I’ve never tested both side by side but I trust Squig’s comment with highlight rolloff being better on the 5D.

In the end, doesn’t matter much... get the camera that gets you excited and create something with it... there are plenty of others that can argue stops of DR or SNR or SNL...

[webrunner5]

Yeah I think I will take the 5D mk III ML output over the BMPCC any day. Now this new 4K BMPCC with a Speedbooster on it, now that may be a lot different story.

What a gift that camera could be to all of us if it pans out like we hope it can.

If it does I will put Grant Petty on my Christmas Card mailing list LoL. May even slip a Dollar Bill in it. Might even start drinking Fosters Beer. Well maybe on that part. ?

[zerocool22]

Yeah I really liked the Image from the 5D III RAW, but the workflow is something else though and I also missed decent 1080p 60fps. Also the screen of the 5D III is a tad too small to make critical checks(never got it working properlly with a external monitor, which also defeats having a dslr type body). 

The pocket 4K will def replace my 5D III for video purposes. 

[Gregormannschaft]

Just dropping here, have we actually seen any decent footage from this camera yet? I didn’t like those staged set tests from NAB earlier this year. I want to see it out in the wild, moving.

[Anaconda_]

Nope. Still waitng, and with only 2 months until release, I don't imagine we'll be waiting much longer.

[kye]

1 minute ago, Anaconda_ said:

I don't imagine we'll be waiting much longer.

Excellent...  The next 100 pages will be hard to fill otherwise