PC4K: ProRes vs. RAW

Hi there,

Has any of you done any ProRes vs. RAW comparisons with the new Pocket 4K? I usually shoot RAW, but in low light the ProRes looks much cleaner, which makes me wonder, wether I should put the effort in denoising at all...

Also interesting this comment I found from Cinema 5D, which implements the the ProRes flavour got updated in the Pocket 4K:

Actually, we got a lot better DR results on ProRes with this camera as opposed to RAW. We reached out to Blackmagic Design for an explanation and received a satisfying answer, and Gunther even had a few calls with them about this to explain.

QUESTION by Gunther to Blackmagic Design:

We have recently tested the BMPCC4K dynamic range using a Xyla DSC labs 21 transmissive chart, and we stumbled across a strange behaviour of the camera:

- we tested in ProRes 4K DCI 25p and got 11.6 stops for a signal to noise ratio of 2, and 12.7 stops for a signal to noise ratio of 1 at ISO 400 (10.5 and 11.8 stops for ISO3200)
- when we tested the same DCI 4K 25p RAW (1:1) we found that the dynamic range reading was lower (by about a half stop). Only when we selected the "highlight recovery" option in DaVinci Resolve and then exported frame grabs into IMATEST we got similar results to ProRes.

This is a bit strange, because we tested also the old Pocket Cinema Camera and it had the same dynamic range reading in ProRes and RAW - and when we enabled the "highlight recovery" option in Resolve, we actually got about 0.6 stops higher DR in RAW that with ProRes.

How come that the DR of the new Pocket 4K is lower in RAW, and only matched ProRes when "highlight recovery" is selected in Resolve? Is ProRes doing this HL recovery automatically?

ANSWER from Blackmagic Design, Andy Buckland:

The Pocket Cinema Camera 4K uses our new demosaic algorithm developed with Blackmagic RAW, before encoding to ProRes. This new algorithm helps to reduce noise and improve the dynamic range in ProRes capture. With Cinema DNG RAW, the demosaic is still performed in DaVinci Resolve. That enables full control and manipulation over debayer settings, noise reduction, and highlight recovery to increase the dynamic range measurement beyond that of ProRes.

By applying minimal temporal noise reduction to the Cinema DNG RAW files in DaVinci Resolve - the lower steps on the waveform can be cleaned up nicely, bringing the available dynamic range of RAW in line with those seen in your ProRes tests. Then by enabling highlight recovery, the available dynamic range in Cinema DNG RAW increases beyond that of ProRes.

ProRes users benefit from the new debayer algorithm in terms of dynamic range and reduced noise at capture, resulting in a faster turnaround.
RAW users benefit from the full control and manipulation of the image, to decide which parameters they would like to tweak and finesse.

What are your thoughts / results?

Cheers

Janis, quite early after the release of the BMPCC4K, someone did a green screen test of a subject, not a test for dynamic range, that showed ProRes doing very well against raw. That explanation from BMD is helpful to explain what’s happening now. It also resulted in much less moiré in a portion of a dark grey synthetic material.

Another couple of good reasons for shooting ProRes when you need a fast turnaround.


Sent from my iPhone using Tapatalk

This would comport with my experience .
I'm really impressed with the improved ProRes image compared w/ other BM cams - really impressed.

I love what they've done with Resolve but it's just so easy/fast to turn ProRes projects around in FCPX - and the new 4.4 update brings NR.

From a practical standpoint, I don't see any compromise shooting ProRes HQ vs raw, but there are real advantages in production.

Janis, quite early after the release of the BMPCC4K, someone did a green screen test of a subject, not a test for dynamic range, that showed ProRes doing very well against raw. That explanation from BMD is helpful to explain what’s happening now. It also resulted in much less moiré in a portion of a dark grey synthetic material.

Another couple of good reasons for shooting ProRes when you need a fast turnaround.


Sent from my iPhone using Tapatalk

Personally, I find the the CDNG to yield a superior image with a lot more flexibiity in post. I also believe that it also shouldn't have the problem of lost recordings from the battery running out.

David, I agree in general with the exception of dealing with moiré as previously noted. Still really looking forward to using Q0 and Q5 and shooting 4K ProRes 422 in camera if I have the media capacity to transcode to 2K ProRes 444 in post. Very good options depending upon a given requirement and your goals.


Sent from my iPhone using Tapatalk

The whole Cosmos sci-fi feature movie was shot in ProRES LT (with a BMPCC). The two creators said that they did tests, and it looked just fine for what they needed. They might have used HQ for green screen only. It indeed looks great. So I think we're splitting hairs here, unless you really need to get back lots of highlights, in which case RAW is best.

If someone chooses to record LT, that’s fine for their needs, I cannot be concerned. It’s their decision.

One of the earliest camera tests I did ran through all the flavors of ProRes on the 4.6K sensor shooting a complex scene. Every single flavor is subtly different. That’s not to say you can’t use any of them. Especially when editing and coloring, everything is easier working with 12bit ProRes so that’s what I shoot when I can.


Sent from my iPhone using Tapatalk

Yep. Remember what I was saying about BRaw in the other thread. So sad it was done this way. It looks like temporal noise removal and image fixes aren't preferenced, which would give you better average dynamic range, better image retention and compression to image retention ratio, and highlight recovery. But what would I know compared to all the stalking image "experts". Seeing they don't see.

Once you sacrifice image detail in intraframe spatial noise removal, it's to a degree, GONE from the image.

New Sony sensor technology coming out on the A7111s. So maybe the next pocket will have similar technology, making the situation better again, as the lattitude may increase and the noise floor may fall lower, with 14bits+ (surprised the current sebsur doesn't have 14 bit mode). Maybe it will be 4.6k-6k 14 bits pocket in nab 2020.

Maybe a good idea to live for today.
The time span between original Pocket and Pocket 4K was five years.
As 2018 draws to a close, many pre-orders remain unfilled.
I think the Pocket 4K will only arrive with a larger number of customers in 2019. You're semi-predicting a 3rd generation Pocket in 2020?
This (becoming expected) relentless churn in technology is horrendous. Let's appreciate what we have and actually use it without discarding it every two or three years.

Chris, you can keep yours until it drops dead for 10 years. But I don't want to own this pocket for 10 years. I don't think the sensor is as good as it could be, I think it could be a generation in advance already, and two generations with the new sensor tech, which might put it where I want to be long term. Sony announced they were reserving new sensor technology for themselves for a year, a few years back. Which means they may release this tech late next year or early the year after, for others to use. For BM its easy, buy the latest off the shelf sensor and accommodate it in the existing design for little extra cost per unit. The we don't have to wait 5 years for a micro as good as the mini 4.6k. The ursa should be 6k-8k next year.

Here is my two cents after about a month of testing...

CDNG...
Understandably requires more work in post to get the noise levels similar to prores. There is obviously more flexibility and the image is generally sharper, better at whatever after enough work. It does not have this interesting "striping" artifact that I have seen in prores when lines such as wires, high contrast lines on house siding, etc are visible. I'll explain below in the prores section.

Prores...
Outstanding IQ in 422 and HQ. I am going to be using the standard 422 flavor for most things. The recovery in highlights is almost the same as CDNG in my testing save the clicky box. You have a lot of flexibility and in a properly exposed shot (I usually overexpose a little to get cleaner images...) I don't think there is that much a difference in the utility of 422 when compared to CDNG unless you are in extreme cases, like highlight clipping in clouds or a grossly incorrect white balance in camera, stuff like that. CDNG is better, but not THAT much better for my purposes as I will be rendering to HD for my deliverables.

I do not shoot "films" though, as in narratives, so I may not be the best person to say what will give appropriate quality for that.

Caveat...
Prores DOES have an interesting striping that happens though on thin poles, wires, house siding, etc. which is really weird. I haven't seen anybody talking about it and it looks to me it is in the demosaicing done in camera. You can see it for example in the linked video below. CDNG does not do this. Generally in a final render with a normal chroma noise reduction, this can be rendered mostly invisible, but it is there. If anybody else has seen this in another camera and knows a trick to removing it, I would love to know how to do so.

My consensus is that I personally will not have much of a reason to shoot in raw. The prores IQ buts me in a good spot to much more quickly get any noise taken care of and the lattitude in post is very good. I am much looking forward to BRAW as I think that will provide an even better compromise between space and IQ.

My understanding is a vast majority of network television shoots Prores even on cameras capable of shooting raw - the raw workflow is just too heavy while the Prores workflow meets all their needs.

That being said, I'm extremely excited about Blackmagic raw coming to the P4k - I think that will be a further game changer to an already amazing camera.

My consensus is that I personally will not have much of a reason to shoot in raw.

I like a consensus of one
And yeah, I'm impressed with how much improved and more malleable the ProRes image is with the P4K.
Shooting HQ in UHD to SD cards is a sweetspot

And yeah, I'm impressed with how much improved and more malleable the ProRes image is with the P4K.
Shooting HQ in UHD to SD cards is a sweetspot

Which sd card do you use to record 4K PRHQ ?
I'm having hard time recording 4K 50p PRLT in sd card

Hey man, it was a split decision, but the consensus of one has spoken.

Which sd card do you use to record 4K PRHQ ?
I'm having hard time recording 4K 50p PRLT in sd card

SanDisk Extreme Pro 95MBS V30.

Have only shot UHD with them.

Here is my two cents after about a month of testing...

CDNG...
Understandably requires more work in post to get the noise levels similar to prores. There is obviously more flexibility and the image is generally sharper, better at whatever after enough work. It does not have this interesting "striping" artifact that I have seen in prores when lines such as wires, high contrast lines on house siding, etc are visible. I'll explain below in the prores section.

Prores...
Outstanding IQ in 422 and HQ. I am going to be using the standard 422 flavor for most things. The recovery in highlights is almost the same as CDNG in my testing save the clicky box. You have a lot of flexibility and in a properly exposed shot (I usually overexpose a little to get cleaner images...) I don't think there is that much a difference in the utility of 422 when compared to CDNG unless you are in extreme cases, like highlight clipping in clouds or a grossly incorrect white balance in camera, stuff like that. CDNG is better, but not THAT much better for my purposes as I will be rendering to HD for my deliverables.

I do not shoot "films" though, as in narratives, so I may not be the best person to say what will give appropriate quality for that.

Caveat...
Prores DOES have an interesting striping that happens though on thin poles, wires, house siding, etc. which is really weird. I haven't seen anybody talking about it and it looks to me it is in the demosaicing done in camera. You can see it for example in the linked video below. CDNG does not do this. Generally in a final render with a normal chroma noise reduction, this can be rendered mostly invisible, but it is there. If anybody else has seen this in another camera and knows a trick to removing it, I would love to know how to do so.

My consensus is that I personally will not have much of a reason to shoot in raw. The prores IQ buts me in a good spot to much more quickly get any noise taken care of and the lattitude in post is very good. I am much looking forward to BRAW as I think that will provide an even better compromise between space and IQ.

Without watching the over 10 minutes of that comparison, I would say that it is probably the internal debayer of the Raw image. You need a better debater algorithm to get rid of it, but then you might loose detail. Processing the image to blur it out instead, would loose detail. If you look at the various cameras there, you see big differences in how it handles this, and how that affects the quality. In a camera without low pass filter, it's worse. The problem is how raw itself works.

Wayne, I tried to provide the time code share but it didn't let me. It's where he zooms in on all of the video files and it can be easily found by scrubbing. If it's a problem with the debayering, it's in the camera as the raw files do not exhibit the phenomenon. If there is something wrong with the debayering in camera I would hope bmd would be looking into fixing this since like I said, it is not visible in raw. This particular issue is the only thing that might hold me back from shooting prores in certain situations.

That's OK. Its a bit complex. One, an fast and easy debayer is going give you stuff. Normal cameras have RGB pixels or and optical low pass filters to smooth out dome issues. But when you have neither the solutions get more complex, or nasty low end processing to further blur out and or nock our the nasties. You see some of the cameras there seem to present some cheap and nasty solution making hazy video. Putting that aside, there are more complex debayerjng algorithms that use many times more processing power, that deal much better with the nasties, but a bit like cheap and nasty, defending on the algorithm. But there is a better way, and as nobody is paying me to reveal it, from my work decades ago, I'm not saying, but will take you closer to a perfect picture.

Now, the issue is that Braw is said to advance things and use a new debayering mechanism. But I'm yet to see how it doesn't loose information which gets you more compressible files. DNG might give you screwed up artifacts, but that is what you get with a Bayer sensor without low pass filter, THAT IS YOUR SENSOR DATA, worts and all. You then have the opportunity to do as good as you can to determine and restore correct detail and beat Braw. It's only false as in, your sensor has limits, but its the best you can work with. Braw, seems something good for recording 8k to deliver 2k from it, but I would have to see more to determine if that is strictly true. This is the strategy used byna number of cameras, Canon cinema, Nokia phones, use higher resolution to produce better quality. Wherever Braw is good for 4k to 2k, I don't know. Dropping resolution, eventually the processing to get rid of noise and artifacts gets close to the downscaled pixels size, and less than it. But things can be so widespread on processing schemes and when you loose real contrast from processing, you loose some pixel accuracy.

Which sd card do you use to record 4K PRHQ ?
I'm having hard time recording 4K 50p PRLT in sd card

SanDisk Extreme Pro 95MBS V30.

Have only shot UHD with them.

Thanks Jim,

Can you shoot UHD 50/60p with your card (prores LT at least) ?

Chris, you can keep yours until it drops dead for 10 years. But I don't want to own this pocket for 10 years. I don't think the sensor is as good as it could be, I think it could be a generation in advance already, and two generations with the new sensor tech, which might put it where I want to be long term. Sony announced they were reserving new sensor technology for themselves for a year, a few years back. Which means they may release this tech late next year or early the year after, for others to use. For BM its easy, buy the latest off the shelf sensor and accommodate it in the existing design for little extra cost per unit. The we don't have to wait 5 years for a micro as good as the mini 4.6k. The ursa should be 6k-8k next year.

The sensor is obviously bad and have no justified future. Just take a look at the horrible skin tones it produce. Almost look like its shoot on that awful Venice camera from sony that nobody cares about.

Yes, I'm shocked, even under "Controlled lighting" look at that horrible burnout on her forehead. Oh, how will it ever cope out in the strong sunlight like I was actually worried about, and to think, according to you they won't need to make a better sensor, even the Venice! Quick ring that Sony factory and tell them to stop making Venice cameras because they are wasting their time, and ring Red too, tell them both to close their factories and businesses, because everybody only needs a Black Magic Pocket 4k video camera according to what you have pointed out, oh woe are us. Of course, a real professional knows that controlled lighting shows nothing about a cameras limitations in feild work. But of course you aren't trying to use that as an example Chris?