BMPCC4K - BRAW Best Format

As with most things, there’s no blanket rule that will cover every situation.

I recently shot interviews at 12:1 and the quality was great. I didn’t have unlimited drive space for this and needed to shoot well over an hour. I’d say 8:1 and 12:1 look about the same.

If you are shooting a giant landscape with tons of detail, you will see a difference between. 12;1 and Q0. For situations like that, I’d do Q0 or a higher constant nitrate. But it will require testing for people to make their own decisions. I know plenty of shooters that love 12:1 for people/faces no matter what. I don’t think I know anyone shooting Q0, but it’s there for people who need it.

With CinemaDNG no longer offered on cameras that support BRAW, that’s a good question. The guidance from Grant Petty seemed to be Q0 or 3:1 to commercial work and possibly Q5 or 5:1 for indie films. We just need a lot more experience shooting a variety of subjects to know which is better overall including a consideration of storage.

I think we will agree with David’s observation that a higher compression ratio can actually look better for people.

No doubt I’ll do the comparison tests anyway, but I’m not going to be surprised if I settle on Q5. That storage requirement for Q5 would be roughly as much as 15% of uncompressed raw or as little as 5%! I’m thinking in practice it will be 10-12% and that’s very nice knowing that the detail will be acceptable.


Sent from my iPhone using Tapatalk

With CinemaDNG no longer offered on cameras that support BRAW, that’s a good question. The guidance from Grant Petty seemed to be Q0 or 3:1 to commercial work and possibly Q5 or 5:1 for indie films. We just need a lot more experience shooting a variety of subjects to know which is better overall including a consideration of storage. I think we will agree with David’s observation that a lower quality can actually look better for people. No doubt I’ll do the comparison tests anyway, but I’m not going to be surprised if I settle on Q5. That storage requirement for Q5 would be roughly as much as 15% of uncompressed raw or as little as 5%! I’m thinking in practice it will be 10-12% and that’s very nice knowing that the detail will be acceptable.

I never said a "lower quality" for people. What I meant was that 8:1 and 12:1 work well for people/interviews because there isn't as much detail in the frame compared with a mountain scene with tons of trees/leaves. Most of the time, people have a background that's out of focus, so the datarate would really be about the face/person. To me, 3:1 does look different (when you pixel peep a busy scene), but that's probably related to the amount of detail in specific scenes which might not be the norm.

I’ll correct my previous post.


Sent from my iPhone using Tapatalk

I just got the BMPCC4K and coming from RED.

Will I need the CFAST for BRAW? Or can I use lower cost compact flash memory cards?

I currently have 2 angelbird AV PRO CF.

Fine. Look at the list of approved cards. I believe your card is there.


Sent from my iPhone using Tapatalk

Angelbird AV PRO CF cards are cf 2.0 compliant.

You’re fast mico, I just noticed they were on the list. CF apparently does not mean Compact Flash for Angelbird.


Sent from my iPhone using Tapatalk

These are quite drastic findings:
https://www.bmpcc4k.tech/2019/03/05/fai ... mware-6-2/


Here are two 1:1 image crops for comparison:

cinemadng-3_1.png (286.22 KiB) Viewed 26873 times

CinemaDNG 3:1

BRAW-3_1.png (197.02 KiB) Viewed 26873 times

BRAW 3:1

The difference in the resolution of textures remains quite visible even when you downscale the image from 4K to 2K - even when you downscale from 4K to SD...

This is a drastic loss of optical resolution.

These are quite drastic findings:
https://www.bmpcc4k.tech/2019/03/05/fai ... mware-6-2/


Here are two 1:1 image crops for comparison:

cinemadng-3_1.png

CinemaDNG 3:1

BRAW-3_1.png

BRAW 3:1

The difference in the resolution of textures remains quite visible even when you downscale the image from 4K to 2K - even when you downscale from 4K to SD...

This is a drastic loss of optical resolution.

Huge difference. It looks like the BRAW image is out-of-focus.

He applied nodes on gamma and extended video to the timeline. Lets see without. Maybe the effects on dng and Braw are different.

A lot of what you see with DNG is false detail (please read about the Shannon/Nyquist limit).

This is Braw Q0 without an OLPF:

https://www.dropbox.com/s/s24mxreh57xy6fx/Braw_ohne_OLPF.png?dl=0

And this is DNG with an OLPF (Rawlite):

https://www.dropbox.com/s/32k69wow5peauvl/DNG_mit_OLPF.png?dl=0

Personally, I dislike fake detail as much as fake news.

cDNG vs BRAW Q0 looks ok as for me. Some aliasing is still visible at both samples but cDNG just produces harder pixel edges

So.. is it mostly the pixel peeping that apparently makes it look not as good? When I look at the images as a whole, there is hardly any difference. So is the detail that might be missing from braw vs cinemaDNG only when you zoom way in?

I will say that despite my wanting the absolute highest quality for no reason at all, the majority of people watch this stuff on phones from youtube. As Peter McKinnon says, HD is still plenty good for just about any project today. Granted I realize 4K and 8K are better for post for some things, and CDNG may be better (lossless) vs BRAW Q0 for some uses, but I would suspect in a large portion of uses, it wont matter.

On another note.. is BRAW 16bit or 12bit? Loaded up in Resolve it says 16bit under bit depth? Is that an error or is it actually recording 16bit video?

So.. is it mostly the pixel peeping that apparently makes it look not as good? When I look at the images as a whole, there is hardly any difference. So is the detail that might be missing from braw vs cinemaDNG only when you zoom way in?

I will say that despite my wanting the absolute highest quality for no reason at all, the majority of people watch this stuff on phones from youtube. As Peter McKinnon says, HD is still plenty good for just about any project today. Granted I realize 4K and 8K are better for post for some things, and CDNG may be better (lossless) vs BRAW Q0 for some uses, but I would suspect in a large portion of uses, it wont matter.

cDNG is sharper but that is kind of fake sharpness that mostly produce aliasing. Also tiny sharpness increase is useless for high resolution 4k sensors, especially in motion. Hope if they port BRAW to BMMCC they adjust it to be a little bit sharper, because a 1080p sensor at 1:1 pixel level de-bayer sharpness is still matters.
BRAW is Y’CbCr codec so it really lacks true RGB sampling usable for something like green screen VFX.

Here is compare test by Frank Glencairn from bmcuser forum:

Meanwhile, on a very sub part machine I use for work, cdng takes 25 minutes to render a sequence. Same shots in compressed braw, 2 minutes. IQ between the two when viewed normally, about identical to my eyes. When not played back to back, would you ever know? Not sure.

Is cdng sharper? Yeah, and I have to denoise it. After fiddling I can get sharper results without compromising too much, and this increases load significantly so I have to cut in HD or turn off grades for smoother playback.

Is braw softer? Yeah, and it's much cleaner out of box, faster to cut, faster to grade, faster to render, has smaller file sizes, easier to manage files...

Honestly I'm a workflow person and the workflow with braw is quite good for this and is a dream compared to CDNG. I'm quite happy with what I'm seeing so far.

Have to agree with you Dune.. and I suspect *most* people are going to drop their jaws when they see the "almost as good" quality but the insane file sizes, speed in work flow, etc. That you can buy a cheap 95mb/s SD card and record 12bit 444 4K30 or so RAW file is just unreal. So looking forward to using this for everything now!

The 12:1 raw looks surprisingly good. I'm sure there are situations where it won't be the best codec to use, but the quality at the compression rate is really amazing. I remember seeing some amazing stuff with 12:1 braw from the Ursa and just could not believe my eyes.

I haven't done a thorough enough analysis of these codecs as they just came available to the pocket 4k today but I really can't wait. 12:1 is looking like my talking head codec if it holds up like what I saw out of the Ursa.

So I typically record in ProRes 4K/30 using LT or 422. Will the 12:1 hold up on part with 422? Or is it not as good? I would assume it would be better since it is 12bit RAW vs 10bit ProRes. I dont have the time to do some of the pixel peeping some of yall are doing. I also dont shoot anything of importance yet, though I did really want to look at shooting stock footage. I am wondering if there will be a .braw stock footage market now... or if they wont allow that.. which would be disasterous in my opinion since if its as good if not better than ProRes/DNxHR, the sites that offer those formats would be crazy not to support .braw format.

I'm shooting BRAW on the MiniPro for quite a while now - I only use Q0. I can't see any reason for using anything less, than the best Quality BRAW I can get for my work. The file sizes of Q0 are already a joke, compared to DNG, so I don't care for even more compression.

Thanks Frank, this is kind of my take on this also.
Cheers

Agreed. Funny Grant suggests Q5 for indie work.
I already use RAW for my DIY 3 minute short films. So Q0 seems like a no brainer.

So I typically record in ProRes 4K/30 using LT or 422. Will the 12:1 hold up on part with 422? Or is it not as good? I would assume it would be better since it is 12bit RAW vs 10bit ProRes. I dont have the time to do some of the pixel peeping some of yall are doing. I also dont shoot anything of importance yet, though I did really want to look at shooting stock footage. I am wondering if there will be a .braw stock footage market now... or if they wont allow that.. which would be disasterous in my opinion since if its as good if not better than ProRes/DNxHR, the sites that offer those formats would be crazy not to support .braw format.

Braw is superior to prores from my experience. I would not hesitate using 12:1 instead of 422. But as others mentioned q0 is less taxing than people believe. The glass has to be super sharp and the picture need to contain tons of moving fine details before it get big in file size. Q5 is a pretty good compromise if you got slow media to record to.

Q0 can go about one third above 3:1 with very detailed and contrasty scenes and down to 5:1 with less demanding image content. Q5 is more or less beginning where Q0 ends at the lower end.

So Q0 is ultimately the best one to use, vs 3:1? Like when would you use 3:1 vs Q0? Green screen work? Or is Q0 fine for that too?

Speaking of green screen, the few saying its not as good as CDNG.. why? I mean I did green screen (for fun) years ago on h.264 file. It worked decently. Not nearly as good as say ProRes or CDNG, but it was workable. I am a hobbyist (as I always say) so some of you are obviously in the big film industry working with the big boys, so I dont know how well it would work there, but I assume Q0 or 3:1 would be just fine for green screen work?

On another note.. is BRAW 16bit or 12bit? Loaded up in Resolve it says 16bit under bit depth? Is that an error or is it actually recording 16bit video?

All BRAW is 12bit log that unpacks to 16bit in Resolve.


Sent from my iPhone using Tapatalk

So Q0 is ultimately the best one to use, vs 3:1? Like when would you use 3:1 vs Q0? Green screen work? Or is Q0 fine for that too?

Speaking of green screen, the few saying its not as good as CDNG.. why? I mean I did green screen (for fun) years ago on h.264 file. It worked decently. Not nearly as good as say ProRes or CDNG, but it was workable. I am a hobbyist (as I always say) so some of you are obviously in the big film industry working with the big boys, so I dont know how well it would work there, but I assume Q0 or 3:1 would be just fine for green screen work?

I used prores 10bit 422 for years with green screen and it was not a big problem at all, and 4K just makes it even easier to get a good key. I have no idea what could be so wrong with BRAW where it warrants the amount of complaining I have seen but, I don't know what they are expecting or shooting. If you light green screen properly, you get a good key. I used to work in VFX for the biggest AAA features and even with the best cameras in the world, you always had people manually rotoscoping issues.

CDNG might technically be worse with green screen compared to BRAW. And of course, with any understanding of the technical aspects to anything, you have to know how to apply that to real life and determine if it matters for you. In real life will it matter? With my experience, and understanding the pocket 4k is a 1300 dollar camera being compared to Arri and Red, no it doesn't.

Dune.. exactly! I mean like I said I was messing around years ago on older AE software with their basic no plugins keying stuff, with a very badly lit green screen (one of those cheap Amazon kits with crap lighting), a crappy $150 HD handycam, and it worked decent. Yah, I could see some green around my kids still, but for the purpose it was fun. That today the software is infinitely better, and the hardware is 2x infinitely better (at least for me vs what I used), I am quite confident pulling a key would be very easy today with BRAW, one that could look quite pro quality. Not that I want to harp too much on this, just that I am overly excited about the potential of this, and want to make sure I understand why it may not be able to do something.

Cant believe I can record up to an hour of 4K 60 DCI Q0 on my cheap $100 512GB SSD right now! I am sure if the scene were move detailed (not just an empty wall), it would be a lot less, but damn, this is insane! I seriously want to sell some old tools and crap, find a way to pull in a few K and buy 2 more of these cameras! This is just unreal how powerful this little thing is.

On another note.. is BRAW 16bit or 12bit? Loaded up in Resolve it says 16bit under bit depth? Is that an error or is it actually recording 16bit video?

BRAW is 12bit log encoded data which, is exactly the same way that ARRI handles ARRIRAW as it is also 12bit log encoded

According to ARRI, their log encoded 12bit ARRIRAW transforms back to 16bit linear without loss, and I think it's safe to assume that it's the same case with BRAW.

So, does that mean the sensor data coming off is 16bit, the same as RED and others, and thus we are essentially getting 16bit video, vs 10bit ProRes? I always thought it was just 2 bits more data, e.g. 12bit RAW vs 10bit ProRes. 16bit is quite a bit more data, if that is the case.

Recorded BRaw with 95/MB’s 128gb as on 3:1 24fps and it worked fine. Also as a vfx as long as you lit your green screen correctly you shouldn’t have a problem especially with BRaw

So, does that mean the sensor data coming off is 16bit...

Not necessarily as I don’t think we’ve been told what the sensor is sending out to the image processor for the BMPCC4K. It could be 16bits but...

You may recall we were told years ago BMD cameras actually took 22 bits (11 bits from two sensor circuits) and that was truncated to at least 16 bits in camera and then output as 12 bit log. By comparison ARRI worked with 28bits internally (14 bits from two sensor circuits).




Sent from my iPhone using Tapatalk

FWIW, shot some tests in all codecs yesterday and ProRes HQ was pretty much equivalent to QO in terms of detail and noise. 12:1 was pretty noisy shooting a forest scene in daylight - ProRes noticeably better.

Bottom line is the that ProRes in the P4K is better done than in any camera I've used from BM. They've come very close, in a practical sense, to raw with what seems to be the same or similar debayering in camera.

It's all good.

I now believe that the previously posted action figure test shots had focus pulling issues (and therefore should be taken with a grain of salt).

Just made my own test pics, with the following results (1:1 crops):

DNG vs BRAW Q0.png (785.04 KiB) Viewed 26322 times

DNG vs BRAW Q5.png (785.06 KiB) Viewed 26322 times

DNG vs BRAW Q0.png (785.04 KiB) Viewed 26322 times

The scene was shot at daylight WB at 400 ISO with a Voigtländer 25mm/f0.95 at f5.6, with the material imported into Resolve and the images captured as screengrabs and exported as PNGs from Resolve's Gallery in the "Color" tab.

Thanks Johan. The only real difference I can plainly see in those crops is that the reds are more muted in the new color science, which I had thought when doing a quick look yesterday. The reds were punched a bit high in the previous color science to where they bled a bit out of the box when shooting saturated reds and now they look like they are more under control, which is a good thing to me.

It looks as many of us initially thought, that the Superman post suffered from the first to report, least correct info situation. Much ado about nothing it seems.

The only real difference I can plainly see in those crops is that the reds are more muted in the new color science, which I had thought when doing a quick look yesterday.

There is a difference in detail resolution between DNG and BRAW, although it's less dramatic than the action figure shots suggested.

To illustrate it with a 200% enlarged crop:

DNG vs BRAW Q0 - 200% crop.png (106.76 KiB) Viewed 26262 times

The only real difference I can plainly see in those crops is that the reds are more muted in the new color science, which I had thought when doing a quick look yesterday.

There is a difference in detail resolution between DNG and BRAW, although it's less dramatic than the action figure shots suggested.

To illustrate it with a 200% enlarged crop:

DNG vs BRAW Q0 - 200% crop.png

I see that now however could not plainly see the difference with the images that were shown previously at the 100% crop. I am also seeing a bit more overall noise in the CDNG, so perhaps a trade off in the demosiacing? Nowhere near the level of Superman thats for sure.

I suspect we will see a lot of examples in the coming days/weeks. I saw a few responses on youtube regarding going to braw loses quality as well. Seems everyone is quick to shout out their opinions. To me as some of you, the workflow benefits trump everything else. I am stunned at how fast this edits in Resolve. Its like going from a dual core laptop to a beastly editing rig. I would be curious to see if older laptops that people have been saying dont work well with Resolve could work well enough with braw footage. Though doubt all of them have the ability to buy one of the 3 cameras that can record braw at this point.

The only real difference I can plainly see in those crops is that the reds are more muted in the new color science, which I had thought when doing a quick look yesterday.

There is a difference in detail resolution between DNG and BRAW, although it's less dramatic than the action figure shots suggested.

To illustrate it with a 200% enlarged crop:

DNG vs BRAW Q0 - 200% crop.png

If seeing an "issue" requires pixel-peeping a still image at 200%, is it a meaningful issue?

Do you think a moving image that has gone through even the very best broadcast-grade compression would be perceptibly better?

I didn't see or state any issue, but only pointed out a difference.

you can keep doing more with RAW files; whereas, with ProRes, there's no real opportunity to do more.

Obviously that's the distinction between raw and any other codec, but my point is how close the new implimentation of ProRes is to the various flavors of Braw.
And there is certainly an opportunity to do a lot "more" with a well shot ProRes image.
That's the point - you can CC it quite a bit without breaking it. And whie you obviously can't adjust ISO in post you can certainly move the WB to a reaonable degree and grade the hell out of it.

The bottom line is if you shoot well (like we always had to) it's virtually impossible to differentiate a well balanced and CCed Prores clip from a Braw clip in any practical sense.

The bottom line is if you shoot well (like we always had to) it's virtually impossible to differentiate a well balanced and CCed Prores clip from a Braw clip in any practical sense.

Totally true, but I think that perspective misses what's special about BRAW -- the simplified, color managed workflow

If you shoot log (aka BMD FIlm) in ProRes, you have to manually color manage that file through post by grading it, or applying LUTS, at every step -- from monitoring on set, to dailies, to the edit, to the color grade.

With BRAW, especially the v1.3 Grant just announced, you can have all the dynamic range and color the sensor can capture, but set the color management as metadata right in the camera. And that metadata will follow along and automatically control how the file is displayed all the way through post right through to the color grade. No guessing about which LUT was used on set that day, which log curve was used on the file, etc. (which is a current headache I run into with all the versions of Canon Log and Slog).
And, you can quickly and easily change that metadata at any point and store that change in the tiny sidecar file. That's a huge workflow efficiency boost. It's how ARRIWRAW and Red RAW .r3d files work in post. But, with BRAW we get that ability on a $1300 camera AND the files are as easy for the computer to decode as ProRes.

As far as my documentary work goes, its totally amazing and it has streamlined every single job I've done since the Ursa Mini Pro got BRAW. And now that the 4K Pocket has it to, I can have that workflow on my A cam and on my B, C, and gimbal cam/crash cam. I'm never going back to ProRes (assuming I have the choice, of course... sometimes producers force particular workflows)

Totally true, but I think that perspective misses what's special about BRAW -- the simplified, color managed workflow

I left out the part about having a few FCPX stations and how fast they are, being optimized for ProRes. For us it's mostly about not needing to go into the (wonderful but complex) resolve universe.

When things get hectic, like they are now, I've got as many as 5-6 projects in aquisition and as many in edit. The simplicity of shooting and exposing with a LUT preview, dropping the files into FCP and just adding it back is great.
The new Gen4 science and ProRes implimentation just makes these files that much more robust and easily gradeable.

What is this BRAW 1.3? Is that only for Ursa Mini? We just got the 1.2 yesterday, and dont recall hearing about 1.3 in his video. What does it add/fix?

I like q5 for its compromise of quality and file size. And you can use SD cards with it.

There’s definitely more detail in a cDNG file compare to prores and BRAW. But the question is if you need that pristine detail. Ironically it’s a bigger chanse to be a tad out of focus rather than actually presive that extra detail. But i lt May show when you start to blow up the image. Say from HD to 4K. You can go further with cDNG. But if all you do is cropping in 5-15% or so because of image stabilization, than there is no problem using braw. I believe all this buzz will ease down when people see how wonderful braw actually is to work with. There are drawbacks, but they’re not many. The goods outweigh the baddies on this one.

The bottom line is if you shoot well (like we always had to) it's virtually impossible to differentiate a well balanced and CCed Prores clip from a Braw clip in any practical sense.

Totally true, but I think that perspective misses what's special about BRAW -- the simplified, color managed workflow

If you shoot log (aka BMD FIlm) in ProRes, you have to manually color manage that file through post by grading it, or applying LUTS, at every step -- from monitoring on set, to dailies, to the edit, to the color grade.

With BRAW, especially the v1.3 Grant just announced, you can have all the dynamic range and color the sensor can capture, but set the color management as metadata right in the camera. And that metadata will follow along and automatically control how the file is displayed all the way through post right through to the color grade. No guessing about which LUT was used on set that day, which log curve was used on the file, etc. (which is a current headache I run into with all the versions of Canon Log and Slog).
And, you can quickly and easily change that metadata at any point and store that change in the tiny sidecar file. That's a huge workflow efficiency boost. It's how ARRIWRAW and Red RAW .r3d files work in post. But, with BRAW we get that ability on a $1300 camera AND the files are as easy for the computer to decode as ProRes.

As far as my documentary work goes, its totally amazing and it has streamlined every single job I've done since the Ursa Mini Pro got BRAW. And now that the 4K Pocket has it to, I can have that workflow on my A cam and on my B, C, and gimbal cam/crash cam. I'm never going back to ProRes (assuming I have the choice, of course... sometimes producers force particular workflows)

Yes, I totally agree!!! I went out of mind trying all the color management settings to get a streamline post processing with ProRes and I was thinking about switching to cDNG; but hey, BM released BRAW so this is a relief, and a big jump in quality, for me at least.

Oyvind, what you said is spot on in my opinion. MOST of us are NOT going to need that extra little bit that CDNG *might* offer. However, ALL of us will absolutely enjoy the faster editing workflow, the smaller file sizes, and if Q5 is all you need, the ability to record it on cheap SD cards... or at least as a backup in case other options are unavailable (sd died, SSD full, etc). What I love is that faster frame rates you can get with it, and that I can get WAY WAY better quality than H.264/5 on cheap SD cards (or cheap SSD to be honest) and go right in to editing it.

Now only if I could go back and rerecord the 13 soccer games I did in ProRes into BRAW. Would make my current few tasks so much faster!

I was quite sceptical towards braw when it first arrived on the mini pro. I thought it looked soft and unimpressive against cDNG.

But it has grown on me, and I have to admit that is actually superior to cDNG in more ways than it sucks. The HD 120 cDNG will be missed on the pocket4k. But I rather want BM to live and prosper, than end with a fatal lawsuit because of cDNG..

And beside... most clients don’t see ANY difference in the end of the day. Also, they mostly don’t pixel peep. Even 12:1 braw is good enough for most, if you do it right from the beginning.

So, does that mean the sensor data coming off is 16bit, the same as RED and others, and thus we are essentially getting 16bit video, vs 10bit ProRes? I always thought it was just 2 bits more data, e.g. 12bit RAW vs 10bit ProRes. 16bit is quite a bit more data, if that is the case.

May be you not know the math behind it
8bit = 256
10bit = 1024
12bit = 4096
If multiply it
16 million vs 1073 million vs 68719 millions of colors
2 bit more is not just 2bit
Too much people not see the point, it’s not only raw the key, but also the color number to do the file.


Inviato dal mio iPhone utilizzando Tapatalk